Cladistics and Paleontology

[Peter Nyikos]

Cladistics, in its broadest sense, studies groups that are classified into
subgroups via what we set theoretic topologists call "a rank 1 system".
This means that in a Venn diagram of subgroups, two subgroups either
do not meet, or one is contained in the other.

And so, the traditional Linnean system is also a rank 1 system, but
it is not the kind that produces phylogenetic trees via the standard
algorithm, nor was it designed to do so. However, for extant species
it does produce a standard cladistic tree -- or would if it were
not limited by the small number of ranks.

I have been referring to the standard algorithm for going from any rank 1
system to a tree. There is a YouTube illustration of the algorithm
in action; it can be linked from here:

https://search.yahoo.com/yhs/search;_ylt=A0LEVkAWsgdXxEsA.ecnnIlQ;_ylu=X3oDMTEwZHBxMGxyBGNvbG8DYmYxBHBvcwMyBHZ0aWQDBHNlYwNyZWwtYm90?p=how+to+make+a+cladogram
[click on How to Build a Cladogram (YouTube)]

The algorithm can be used to build all kinds of trees having nothing
to do with phylogeny; an example I've used in the past is a rank 1
classification of mountains using various "characters" of individual
mountains, and I've speculated that the "sister group"
of the Matterhorn might be a mountain in the Himalayas.

Anyway, a rank 1 classification system works perfectly for extant taxa,
but it gets worse and worse the further back one goes in time. For instance,
the only taxa containing *Panderichthys* are the genus itself and
a super-class that includes all terrestrial vertebrates.

I'll have more to say about the drawbacks of rank 1 classification
systems for paleontologists in later posts to this thread.

Peter Nyikos
Professor, Dept. of Mathematics -- standard disclaimer--
University of South Carolina
http://people.math.sc.edu/nyikos/

[Peter Nyikos]

This is a reply to a post by John Harshman on the thread,
Subject: Re: Archaeorhynchus spathula
in which I will critique the philosophy that results inevitably
in a rank 1 classification of animals.

On Thursday, April 7, 2016 at 5:51:00 PM UTC-4, John Harshman wrote:
> On 4/7/16 12:00 PM, Peter Nyikos wrote:
> > On Thursday, April 7, 2016 at 1:44:07 PM UTC-4, John Harshman wrote:
> >> On 4/7/16 10:19 AM, Peter Nyikos wrote:

> >>> Oops, I had glossed over something you wrote further down. I'd
> >>> still be astounded if, say, any published cladogram of monotremes listed ALL
> >>> the apomorphies of the few species we know of that have occurred
> >>> since LCA of Monotremata.
> >>>
> >>> Just think of all the differences between *Ornithorynchus* and the
> >>> echidnas.
> >>
> >> Yes, autapomorphies are often excluded,
> >
> > You cannot be sure that most or all are not synapomorphies, with
> > a huge number of side branches for which we lack fossil evidence.
> > Monotremes are an excellent example of this uncertainty.
>
> Of course you can't. But so what? If it's an autapomorphy with regard to
> all taxa you know about, it doesn't help in phylogenetic analysis

It does help in knowing what characters (plesimorphies!) the LCA had, if you
know ALL the automorphies that the extant monotremes have.

Since you are not a paleontologist, I cannot expect you to appreciate
the value of the information about the LCA. A paleontologist would
naturally be excited to find a fossil that has some key [1] characters
and has them, and all other characters of the fossil, match those
of the LCA.

[1] Teeth are the most key for mammals, but I can't expect an ornithologist
like you to appreciate that.

And so, there was great excitement about the first monotreme
fossil from before the Miocene, a Cretaceous fossil of part
of a jawbone with teeth, *Steropodon*. Thanks to the extinct
Miocene *Obdurodon* having retained enough teeth,
there was no doubt about it being a monotreme.

Some popularizers went overboard and called it a "platypus", but
according to the latest estimate of which I know, the echidnas
and platypuses diverged well after the Paleocene.

>(with
> the exception, again, of maximum likelihood methods, in which it helps
> to determine various evolutionary parameters).
>
> >> as are invariant characters. (I
> >> will note that these should not be excluded from maximum likelihood
> >> analyses, and there are likelihood methods that try to correct for their
> >> exclusion.) That's why I talk about "relevant characters". What point
> >> are you wanting to make about that?
> >
> > One point is that "relevant characters" are a piss poor substitute
> > for a thorough analysis of the characters of the LCA.
>
> I'm not sure you know what you're saying here. How would you propose to
> analyze those characters?

I propose to analyze ALL characters, not just those that are subjectively
deemed relevant. For instance, feathers of various maniraptorans are
omitted from cladistic analyses endorsed by you -- I learned this from you.

And this emphasis on characters rather than taxa is due to the
philosophy that leads to the cladistic rank 1 classification.
An illustration is the only comment you made on a thread chock
full of fascinating paleontological information about a whale
fossil:

Doesn't matter if it's an ancestor as long as it preserves
intermediate character states.

In short, you are opting for the trees [characters] and don't
care if the forest [the ancestral species] is missed.

Nothing personal, mind you: you are just expounding on the
way cladists think and work nowadays.

> > Another is that "taxa under consideration" is can be a very unsatisfactory
> > approximation to biological understanding.
>
> That wasn't quite a sentence.

Yeah, that second "is" doesn't belong. What you saw was an imperfect
revision of a sentence that I typed earlier.

> And I'm not sure what you mean there
> either. Neither of these points is discernable in what you said, as far
> as I can tell.

I can't figure out what your problem is here. Are you complaining
because you can't connect the dots between what I am saying here
and what I said earlier? Then let it go, and start just evaluating
what I say on its own merits.

You've used the quoted phrases before. Use your own concept of
what they are about.

Peter Nyikos
Professor, Department of Math. -- standard disclaimer --
U. of South Carolina at Columbia
http://www.math.sc.edu/~nyikos/

[John Harshman]

This is of course only true if the ranks all have names and all ranks
must be used throughout the system.

[John Harshman]

On 4/8/16 8:39 AM, Peter Nyikos wrote:
> This is a reply to a post by John Harshman on the thread,
> Subject: Re: Archaeorhynchus spathula
> in which I will critique the philosophy that results inevitably
> in a rank 1 classification of animals.
>
> On Thursday, April 7, 2016 at 5:51:00 PM UTC-4, John Harshman wrote:
>> On 4/7/16 12:00 PM, Peter Nyikos wrote:
>>> On Thursday, April 7, 2016 at 1:44:07 PM UTC-4, John Harshman wrote:
>>>> On 4/7/16 10:19 AM, Peter Nyikos wrote:
>
>>>>> Oops, I had glossed over something you wrote further down. I'd
>>>>> still be astounded if, say, any published cladogram of monotremes listed ALL
>>>>> the apomorphies of the few species we know of that have occurred
>>>>> since LCA of Monotremata.
>>>>>
>>>>> Just think of all the differences between *Ornithorynchus* and the
>>>>> echidnas.
>>>>
>>>> Yes, autapomorphies are often excluded,
>>>
>>> You cannot be sure that most or all are not synapomorphies, with
>>> a huge number of side branches for which we lack fossil evidence.
>>> Monotremes are an excellent example of this uncertainty.
>>
>> Of course you can't. But so what? If it's an autapomorphy with regard to
>> all taxa you know about, it doesn't help in phylogenetic analysis
>
> It does help in knowing what characters (plesimorphies!) the LCA had, if you
> know ALL the automorphies that the extant monotremes have.

No it doesn't. Why?

> Since you are not a paleontologist, I cannot expect you to appreciate
> the value of the information about the LCA. A paleontologist would
> naturally be excited to find a fossil that has some key [1] characters
> and has them, and all other characters of the fossil, match those
> of the LCA.

I submit that I know much more about what paleontologists want and why
they want them than you do. And why not? Since you're not a
paleontologist, apparently a requirement for understanding, who could
expect you to know anything at all.

> [1] Teeth are the most key for mammals, but I can't expect an ornithologist
> like you to appreciate that.

This is perhaps the silliest thing you have said to date. I do know a
bit about mammals. But I can't expect a topologist like you to
appreciate that.

In short: you dismiss my knowledge on the basis of my profession, not
noticing that your profession is much farther from the subject and also
not noticing the contradiction in claiming superior knowledge on that basis.

> And so, there was great excitement about the first monotreme
> fossil from before the Miocene, a Cretaceous fossil of part
> of a jawbone with teeth, *Steropodon*. Thanks to the extinct
> Miocene *Obdurodon* having retained enough teeth,
> there was no doubt about it being a monotreme.
>
> Some popularizers went overboard and called it a "platypus", but
> according to the latest estimate of which I know, the echidnas
> and platypuses diverged well after the Paleocene.

Very nice, but so?

>> (with
>> the exception, again, of maximum likelihood methods, in which it helps
>> to determine various evolutionary parameters).
>>
>>>> as are invariant characters. (I
>>>> will note that these should not be excluded from maximum likelihood
>>>> analyses, and there are likelihood methods that try to correct for their
>>>> exclusion.) That's why I talk about "relevant characters". What point
>>>> are you wanting to make about that?
>>>
>>> One point is that "relevant characters" are a piss poor substitute
>>> for a thorough analysis of the characters of the LCA.
>>
>> I'm not sure you know what you're saying here. How would you propose to
>> analyze those characters?
>
> I propose to analyze ALL characters, not just those that are subjectively
> deemed relevant. For instance, feathers of various maniraptorans are
> omitted from cladistic analyses endorsed by you -- I learned this from you.

You really don't mean that. You don't, for example, propose to analyze
invariant characters. You don't really know what you're saying.

> And this emphasis on characters rather than taxa is due to the
> philosophy that leads to the cladistic rank 1 classification.

I don't understand the connection. Could you explain?

> An illustration is the only comment you made on a thread chock
> full of fascinating paleontological information about a whale
> fossil:
>
> Doesn't matter if it's an ancestor as long as it preserves
> intermediate character states.
>
> In short, you are opting for the trees [characters] and don't
> care if the forest [the ancestral species] is missed.

Not the case. It's just that we can't (as I have explained many times)
identify the ancestral species. Why go so far beyond the data? What do
we gain? So far you have not stated any actual advantages to your
"ancestor candidate" idea. This all seems driven by nothing more than
your desire to recover the good old days of your youth.

> Nothing personal, mind you: you are just expounding on the
> way cladists think and work nowadays.

You apparently don't realize how condescending you sound throughout this
post. This would be a good place to inform you.

>>> Another is that "taxa under consideration" is can be a very unsatisfactory
>>> approximation to biological understanding.
>>
>> That wasn't quite a sentence.
>
> Yeah, that second "is" doesn't belong. What you saw was an imperfect
> revision of a sentence that I typed earlier.
>
>> And I'm not sure what you mean there
>> either. Neither of these points is discernable in what you said, as far
>> as I can tell.
>
> I can't figure out what your problem is here. Are you complaining
> because you can't connect the dots between what I am saying here
> and what I said earlier? Then let it go, and start just evaluating
> what I say on its own merits.
>
> You've used the quoted phrases before. Use your own concept of
> what they are about.

I don't think my concept and your concept match very well. Yes, I want
you to connect the dots to some point you want to make about something.
What, for example, is unsatisfactory about that approach?

[RSNorman]

You seem to be identifying "cladistics" with "producing a rank 1
classification". It is possible to create all sorts of rank 1
classifications for any arbitrary set (OK, any finite set). Cladistics
requires the classification be rank 1 but that is insufficient. There
must be a particular biological association determining the subset
relation or, if you prefer, the daughter nodes underlying any specific
node in the rooted tree. It is in that association, the ensemble of
shared inherited characteristics, that relates the cladistic structure
of classification with phylogeny, the evolutionary sequence of
relationships.

Talking about "rank 1 classifications" in the absence of the reasons
for producing this specific rank1 classification over that one is to
deny the biological basis of the classificatin.

[erik simpson]

Wait for it. He seems to be saying that rank 1 classification schemes are
inadequate in some way. I'd bet a nickle that overlapping paraphyletic
groups will appear before long.

[Oxyaena]

Linnaean Taxonomy has changed a lot over the years to accommodate
changes in biology, especially the coming of evolution in 1859. Linnaeus
first started out with species, genus, family, order, class, and
kingdom. But it became apparent there would have to be changes, as
relationships between species uncovered "in the light of evolution",
which is why we came up with superfamilies, infraclasses, subclasses,
superclasses, infraphylums, subphylums, superphylums, kingdoms, and domains.

Linnaeus' system works well with extant species, as already mentioned,
to an extent. As one still needs to work out the phylogenetic
relationships, which is easy with extant species because of the
molecular evidence, but with fossil species more than a million years
back in time it becomes impossible, and one has to rely on morphological
evidence, which isn't fail proof, a classic example would be the
Ungulata, with molecular evidence the Paenungulates were moved away into
a different cohort, which left only Perrisodactlys and Artiodactyls to
be descended from condylarths, which itself is a wastebasket taxon.



--
"We are all atheists about most gods humanity has ever believed in, some
of us just go one god further." - Richard Dawkins

http://oxyaena.org/

[Popping mad]

On Fri, 08 Apr 2016 08:18:27 -0700, Peter Nyikos wrote:

>
> https://search.yahoo.com/yhs/
search;_ylt=A0LEVkAWsgdXxEsA.ecnnIlQ;_ylu=X3oDMTEwZHBxMGxyBGNvbG8DYmYxBHBvcwMyBHZ0aWQDBHNlYwNyZWwtYm90?

p=how+to+make+a+cladogram
> [click on How to Build a Cladogram (YouTube)]
>
> The algorithm can be used to build all kinds of trees having nothing
> to do with phylogeny; an example I've used in the past is a rank 1
> classification of mountains using various "characters" of individual
> mountains, and I've speculated that the "sister group"
> of the Matterhorn might be a mountain in the Himalayas.

As I've been working on my masters degree, I've increasingly become aware
how graphic algorithms can be used to determine the relationship between
objects and I'm not certain that the best theories have been applied to
paleontology.

https://en.wikipedia.org/wiki/Dijkstra's_algorithm

[Popping mad]

On Fri, 08 Apr 2016 09:40:40 -0700, John Harshman wrote:


> I don't think my concept and your concept match very well. Yes, I want
> you to connect the dots to some point you want to make about something.
> What, for example, is unsatisfactory about that approach?

CONNECT THE DOTS IS WHAT IT IS ALL ABOUT

[John Harshman]

You may be right, but that particular algorithm doesn't seem to me like
it ought ot be useful.

[Popping mad]

On Sun, 10 Apr 2016 06:22:24 -0700, John Harshman wrote:

>> https://en.wikipedia.org/wiki/Dijkstra's_algorithm
>>
> You may be right, but that particular algorithm doesn't seem to me like
> it ought ot be useful.

It is useful because it allows you to detrerman how closely related two
different objects are, in a crowd of objects.

[John Harshman]

You may have a different definition of "closely related" than a
biologist would. Under the biological definition, "objects" would not be
closely related; only organisms would be.

Based on my limited scan, Dijkstra's algorithm appears to produce a
Steiner tree, if it produces a tree at all, which it apparently will
only sometimes do. Is that true? That just doesn't bear much resemblance
to the actual trees of relationships among species.

[RSNorman]

You are quite right, John, Dijkstra is no help at all.

You don't need Dijkstra's algorithm to discover the distance between
any two objects on a simply connected rooted tree, one without any
closed loops and the only kind useful in phylogenetics. It only is
needed if there are multiple closed paths. And then you still need to
know the distance between any two points connected by a single line,
that is, the length of every edge in the graph.

[John Harshman]

On 4/11/16 4:35 PM, RSNorman wrote:
> On Mon, 11 Apr 2016 16:07:20 -0700, John Harshman
> <jhar...@pacbell.net> wrote:
>
>> On 4/11/16 2:33 PM, Popping mad wrote:
>>> On Sun, 10 Apr 2016 06:22:24 -0700, John Harshman wrote:
>>>
>>>>> https://en.wikipedia.org/wiki/Dijkstra's_algorithm
>>>>>
>>>> You may be right, but that particular algorithm doesn't seem to me like
>>>> it ought ot be useful.
>>>
>>> It is useful because it allows you to detrerman how closely related two
>>> different objects are, in a crowd of objects.
>>>
>> You may have a different definition of "closely related" than a
>> biologist would. Under the biological definition, "objects" would not be
>> closely related; only organisms would be.
>>
>> Based on my limited scan, Dijkstra's algorithm appears to produce a
>> Steiner tree, if it produces a tree at all, which it apparently will
>> only sometimes do. Is that true? That just doesn't bear much resemblance
>> to the actual trees of relationships among species.
>
> You are quite right, John, Dijkstra is no help at all.
>
> You don't need Dijkstra's algorithm to discover the distance between
> any two objects on a simply connected rooted tree, one without any
> closed loops and the only kind useful in phylogenetics.

Note that even a Steiner tree is not a phylogenetic tree, as its
branches connect only what we would call terminal taxa -- it
reconstructs no internal nodes or branches.

> It only is
> needed if there are multiple closed paths. And then you still need to
> know the distance between any two points connected by a single line,
> that is, the length of every edge in the graph.

Also true. And of course in phylogenetic algorithms, the lengths of
branches (the only phylogenetic interpretation of "distance between any
two points") are parameters to be estimated, not data.

[John Harshman]

Oops, wrong terminology. I meant a minimum spanning tree. A Steiner tree
does allow for internal nodes to be created as needed.

[Peter Nyikos]

Harshman knows better. I think you know better too, erik.

Both of you, I believe, know that the absolute adherence to
rank 1 classifications is my main complaint about cladistics,
but I've known about MP, NJ, and PAUP (without really knowing
how they work) since 1997.

<snip Norman preaching to the choir>

> Wait for it. He seems to be saying that rank 1 classification schemes are
> inadequate in some way.

As I already did in the OP, preserved above. Didn't you read it?

> I'd bet a nickle that overlapping paraphyletic
> groups will appear before long.

You lose. I haven't advocated those since 2001, or maybe 2000.

As I'm sure you know, I *have* been campaigning for
parallel classifications, such as are found (in highly abbreviated form)
at the end of the following book, the textbook for the class on
Comparative Vertebrate Anatomy here at the University of South Carolina.

Kardong, Kenneth V.: Vertebrates: comparative anatomy, function,
evolution, McGraw-Hill, 6th ed., 2012.

Peter Nyikos
Professor, Department of Math. -- standard disclaimer --

U. of S. Carolina at Columbia
http://people.math.sc.edu/nyikos/

[John Harshman]

Actually, no, I don't know that. What your main complaint seems to be is
that the rank 1 classifications in cladistic classfication (as opposed
to tree-building methods) aren't allowed to include paraphyletic groups.
Just what's wrong with rank 1 classifications is unclear, because you
have told us yet.

All you have said about that is this:

"Anyway, a rank 1 classification system works perfectly for extant taxa,
but it gets worse and worse the further back one goes in time. For
instance, the only taxa containing *Panderichthys* are the genus itself
and a super-class that includes all terrestrial vertebrates."

That's an example, not a reason, and at any rate I don't see the problem
with that example. What is the problem?

Further, that's a problem with a ranked classification, not a rank 1
classification. Wouldn't your proposed solution still be a rank 1
classification? Unless you propose overlapping groups, which you deny.
Even parallel classifications are just two rank 1 classifications, not one.

>> Wait for it. He seems to be saying that rank 1 classification schemes are
>> inadequate in some way.
>
> As I already did in the OP, preserved above. Didn't you read it?

You said they're inadequate in some way, but you haven't explained what
the way is. Could you explain what the way is? And while you're at it,
you can explain the preferable alternative and what its advantages are.

[Peter Nyikos]

You may not remember that far back on this issue because we had
relatively little to do with each other except where the
cladophile prohibition of paraphyletic taxa was concerned.
But I was an avid reader of your exchanges with Cal King,
who had all kinds of bones to pick with cladistic methods of
establishing phylogeny.

Cal King was someone I'd love to have return to sci.bio.paleontology;
he wasn't much for off-topic stuff or personal attacks.

> What your main complaint seems to be is
> that the rank 1 classifications in cladistic classfication (as opposed
> to tree-building methods) aren't allowed to include paraphyletic groups.

You are conflating two utterly different things here. It would defeat
the main practical application of the rank 1 cladistic classification --
reconstructing phlogeny -- to include paraphyletic groups. That
is also the main reason why I am all for two parallel classification
schemes, and that is why you are completely off base with what my main
complaint "seems to be".

> Just what's wrong with rank 1 classifications is unclear, because you

> have [not] told us yet.

Come off it. I have made it clear that the traditional system
is ALSO a rank 1 system, but was never intended for depicting
phylogeny without the aid of "trees of bubble taxa"-- something
you couldn't even recognize as trees, even though they depicted
phylogeny.

My real "cladistic" beef is the prohibition by
your kind of paraphyletic taxa, except -- very reluctantly -- genera.

One disagreement between you and Cal King centered on his claim
that the king snake genus *Lampropeltis* is paraphyletic, and
that *Getulus* is a daughter taxon.

I don't recall why you disagreed with him on that, and would
like to know why.

[Trivia: Cal King's moniker --he never told us his real name -- was
based on "California Kingsnake," and his (fake ?) e-mail address began with
"getulus".]

Continued in next reply.

Peter Nyikos
Professor, Department of Math. -- standard disclaimer --

[Peter Nyikos]

On Tuesday, April 12, 2016 at 9:55:54 AM UTC-4, John Harshman wrote:

> On 4/12/16 5:07 AM, Peter Nyikos wrote:
> > On Friday, April 8, 2016 at 1:19:03 PM UTC-4, erik simpson wrote:

Repeating a sentence from my first reply, for context:

> Just what's wrong with rank 1 classifications is unclear, because you
> have told us yet.

See preceding reply for refutation of that; it also refutes the
first line below.

> All you have said about that is this:
>
> "Anyway, a rank 1 classification system works perfectly for extant taxa,
> but it gets worse and worse the further back one goes in time. For
> instance, the only taxa containing *Panderichthys* are the genus itself
> and a super-class that includes all terrestrial vertebrates."
>
> That's an example, not a reason,

It is one for anyone who knows the point of having a nested hierarchy:
the use of it for getting more and more similar to a species as
one goes down the hierarchy. You cladophiles typically use a single
blanket adjective "similar" and never attempt to qualify it
in a really useful way.

In this day and age, you could almost do away with cladistic
classification and just refer readers to online phylogenetic
trees.

I say "almost" because you still need names for the clades.
But you could also put up with traditional classification systems
and just use their names for taxa which you believe to be clades.
See below about "Sarcopterygii."

> and at any rate I don't see the problem
> with that example. What is the problem?

The two or three taxa are useless to get a feel for how anatomically
similar *Tiktaalik* is to *Acanthostega*. If I were to bring
the traditional classification up to date, I would put these
two into the same family at the base of a super-class
of semi-terrestrial vertebrates, and supplement it with
a bubble tree linking that family to the sister clade of *Acanthostega*
on the one hand and to the paraphyletic panderichthyan family
in the paraphyletic subclass Choanichthyies [clade Sarcopterygii]
on the other.

That way, y'all get to keep your clade "Sarcopterygii."
Unlike you, I'm not one for commandeering names from one
system to the other.

> Further, that's a problem with a ranked classification, not a rank 1
> classification.

I'd like to see the reasoning behind this amazing claim.

> Wouldn't your proposed solution still be a rank 1
> classification?

It is TWO rank 1 classifications, one for reconstructing phylogeny
and the other for giving the concept of similarity some substance.

<snip talk.origins style sentence>

> Even parallel classifications are just two rank 1 classifications, not one.

So what?

> >> Wait for it. He seems to be saying that rank 1 classification schemes are
> >> inadequate in some way.
> >
> > As I already did in the OP, preserved above. Didn't you read it?

<snip inappropriate questions by you>

Peter Nyikos
Professor, Department of Math. -- standard disclaimer --

Univ. of South Carolina at Columbia
http://www.math.sc.edu/~nyikos/

[John Harshman]

On 4/13/16 8:26 AM, Peter Nyikos wrote:
> On Tuesday, April 12, 2016 at 9:55:54 AM UTC-4, John Harshman wrote:
>> On 4/12/16 5:07 AM, Peter Nyikos wrote:
>>> On Friday, April 8, 2016 at 1:19:03 PM UTC-4, erik simpson wrote:
>
> Repeating a sentence from my first reply, for context:
>
>> Just what's wrong with rank 1 classifications is unclear, because you
>> have told us yet.
>
> See preceding reply for refutation of that; it also refutes the
> first line below.

I don't think you refuted it at all.

>> All you have said about that is this:
>>
>> "Anyway, a rank 1 classification system works perfectly for extant taxa,
>> but it gets worse and worse the further back one goes in time. For
>> instance, the only taxa containing *Panderichthys* are the genus itself
>> and a super-class that includes all terrestrial vertebrates."
>>
>> That's an example, not a reason,
>
> It is one for anyone who knows the point of having a nested hierarchy:
> the use of it for getting more and more similar to a species as
> one goes down the hierarchy. You cladophiles typically use a single
> blanket adjective "similar" and never attempt to qualify it
> in a really useful way.

I see a problem here: we disagree on what the point of having a nested
hierarchy is. I think it's to represent the actual shape of the tree of
life, not similarity.

> In this day and age, you could almost do away with cladistic
> classification and just refer readers to online phylogenetic
> trees.

> I say "almost" because you still need names for the clades.

True. You could just look at trees, and a good way to graphically
represent a classification is to label the tree rather than showing an
indented list. I don't see a problem with that.

> But you could also put up with traditional classification systems
> and just use their names for taxa which you believe to be clades.
> See below about "Sarcopterygii."

You could, but why would you? The traditional classification system is
an unhappy compromise between descent and similarity which satisfies
neither very well.

>> and at any rate I don't see the problem
>> with that example. What is the problem?
>
> The two or three taxa are useless to get a feel for how anatomically
> similar *Tiktaalik* is to *Acanthostega*. If I were to bring
> the traditional classification up to date, I would put these
> two into the same family at the base of a super-class
> of semi-terrestrial vertebrates, and supplement it with
> a bubble tree linking that family to the sister clade of *Acanthostega*
> on the one hand and to the paraphyletic panderichthyan family
> in the paraphyletic subclass Choanichthyies [clade Sarcopterygii]
> on the other.

That seems to lead only to vast confusion, as any system of two mutually
incompatible classifications must. Good luck remembering which
classifications use which names.

Further, even the paraphyletic classification can't achieve your goal of
uniting similar taxa. Given evolution, you must draw arbitrary lines
separating similar species and grouping each with less similar ones.
You might be able to draw the line at a large gap in morphology, while
hoping that no intermediates to fill that gap are ever found. But that
hope seems misplaced.

Here, you arbitrarily separate Tiktaalik from, e.g., Panderichthys and
Acanthostega from, again e.g., Ichthyostega. Why? How does this help
usunderstand evolution in any way?

> That way, y'all get to keep your clade "Sarcopterygii."
> Unlike you, I'm not one for commandeering names from one
> system to the other.

You aren't a systematist, so it doesn't really matter. You will never
have to make such a decision.

>> Further, that's a problem with a ranked classification, not a rank 1
>> classification.
>
> I'd like to see the reasoning behind this amazing claim.

I take it back. I didn't see what you thought the problem was. It's a
problem (though I deny that failure to represent disparity is problem at
all) with any rank 1 classification of life, including the one you
prefer. If you divide reptiles from mammals and from birds, there is no
point at which you can represent disparity in the division: you will
always divide similar species from similar species.

>> Wouldn't your proposed solution still be a rank 1
>> classification?
>
> It is TWO rank 1 classifications, one for reconstructing phylogeny
> and the other for giving the concept of similarity some substance.

That last phrase is quite vague, fortunately so, as a clear statement,
say "use a classification to represent similarity" would show the
impossibility of that goal. Again, your second classification is an
uneasy compromise that works for neither, and in fact there is no rank 1
classification that would represent similarity unless there were big
morphological gaps. That's why it works as well as it does for extant
taxa. For fossils it's completely hopeless from the start.

> <snip talk.origins style sentence>
>
>> Even parallel classifications are just two rank 1 classifications, not one.
>
> So what?

Well, you started this by saying you would tell us what's wrong with
rank 1 classifications. But you ended up telling us that what's wrong is
that you need two of them, not just 1. That doesn't seem to me to be a
flaw with the classification at all.

>>>> Wait for it. He seems to be saying that rank 1 classification schemes are
>>>> inadequate in some way.
>>>
>>> As I already did in the OP, preserved above. Didn't you read it?

I did, but your OP was not at all clear. The introduction of "rank 1
classifications" seems tailor-made for confusion, as it invites
conflation of phylogenetic trees with classifications with the
non-phylogenetic trees you could construct by translating traditional
classifications. And you seem to do some of that conflation yourself. It
may be unintentional conflation, but it makes it hard to understand what
you mean. Further, it turned out that the solution you had in mind
wasn't even mentioned in the OP. Finally, both the problem and its
"solution" rely on a number of assumptions unstated in the OP, many of
which are unintuitive to a systematist and/or wrong.

> <snip inappropriate questions by you>

They didn't seem inappropriate to me, though you seem above to have
answered them, more or less.

[John Harshman]

I'm not sure how this relates. What I don't know is that your main
complaint is the absolute adherence to rank 1 classifications. But
traditional classifications are rank 1 classifications. Nobody at all is
proposing any other sort of classification. Thus I am confused.

> Cal King was someone I'd love to have return to sci.bio.paleontology;
> he wasn't much for off-topic stuff or personal attacks.

Wasn't that just a bit of off-topic stuff?

>> What your main complaint seems to be is
>> that the rank 1 classifications in cladistic classfication (as opposed
>> to tree-building methods) aren't allowed to include paraphyletic groups.
>
> You are conflating two utterly different things here. It would defeat
> the main practical application of the rank 1 cladistic classification --
> reconstructing phlogeny -- to include paraphyletic groups.

I think you mean "representing" rather than "reconstructing". But of
course. That's why we don't like paraphyletic groups.

> That
> is also the main reason why I am all for two parallel classification
> schemes, and that is why you are completely off base with what my main
> complaint "seems to be".

See, it's very confusing that you complain about rank 1 classifications
but your proposal is to use two rank 1 classifications. It's like saying
you hate apples, when what you mean is that you don't like having just
one apple; you want both a pippin and a macintosh.

>> Just what's wrong with rank 1 classifications is unclear, because you
>> have [not] told us yet.
>
> Come off it. I have made it clear that the traditional system
> is ALSO a rank 1 system, but was never intended for depicting
> phylogeny without the aid of "trees of bubble taxa"-- something
> you couldn't even recognize as trees, even though they depicted
> phylogeny.

Bubble diagrams aren't trees. They're something like trees, I'll agree.
That traditional classifications were never intended for depicting
phylogeny is a problem.

> My real "cladistic" beef is the prohibition by
> your kind of paraphyletic taxa, except -- very reluctantly -- genera.

What do you mean by "your kind of paraphyletic taxa"? Is there more than
one kind?

> One disagreement between you and Cal King centered on his claim
> that the king snake genus *Lampropeltis* is paraphyletic, and
> that *Getulus* is a daughter taxon.
>
> I don't recall why you disagreed with him on that, and would
> like to know why.

I have no idea. In modern practice, by the way, we reject paraphetic
genera as much as any paraphyletic taxon.

[Popping mad]

On Mon, 11 Apr 2016 16:07:20 -0700, John Harshman wrote:

> You may have a different definition of "closely related" than a
> biologist would. Under the biological definition, "objects" would not be
> closely related; only organisms would be.

objects is a generic superclass term in most fields, including but not
limited to Mathmatics, Computer Sciences, and Physics.

This is a mathmatical principle. Two objects are clsely related, in this
case species and in the general, the more shared charactorists they
have. This is objectively measurable. That would be the paleontology
part...creating objective charactoristics that can be measured,
described, quantified and notated.

[Popping mad]

On Mon, 11 Apr 2016 16:35:16 -0700, RSNorman wrote:

> Dijkstra's algorithm to discover the distance between any two objects on
> a simply connected rooted tree,

Yeah you do, because it demonstrates the shortest path and in fact there
ARE loops in the data and determining the shortest path is the only means
of determining the trees shap.

[RSNorman]

Perhaps I used the phrase "simply connected" from topology in a way
that does not apply to graph theory.

The biological tree of descent is necessarily acyclic. There are no
loops or cycles in the data.

[John Harshman]

I'm sorry, but this is incorrect in regard to species. It would be true
if evolution were perfectly clocklike, but it isn't. If lineages evolve
at different rates, the two most similar species may not be the two most
closely related. If you want to learn something about how phylogenetics
is done, and why, I could recommend a book or two.

[John Harshman]

Well, there are in the case of allopolyploid species. But that's not all
that common.

[RSNorman]

When you consider the tree to be directed the cycle disappears.

[John Harshman]

True. But very few methods of analysis assume a rooted tree, and almost
all models are time-reversible.

[ruben safir]

On 04/14/2016 12:07 AM, RSNorman wrote:
>> Yeah you do, because it demonstrates the shortest path and in fact there
>> >ARE loops in the data and determining the shortest path is the only means
>> >of determining the trees shap.
> Perhaps I used the phrase "simply connected" from topology in a way
> that does not apply to graph theory.
>
> The biological tree of descent is necessarily acyclic. There are no
> loops or cycles in the data.

The result is, the data is NOT, which is why it is essential to find the
shortest path

which is the essential problem.

It might be that the resulting analysis is not either, because it is an
assumption that probably isn't true. There is no reason to believe that
a single tetrapod came out of the water to spawn all tetrapods. It
could very well be a group of closely relates tetrapods, that might even
have been geographically isolated from each other.

I'm not sure you really understand graph theory after reading the
previous posts.

Ruben

[ruben safir]

On 04/14/2016 11:03 AM, John Harshman wrote:
>>
> True. But very few methods of analysis assume a rooted tree, and almost
> all models are time-reversible.

irrelevant to the analysis. It doesn't make a difference if there is a
rooted tree or not. what matters is the shortest path between data points.

[ruben safir]

On 04/14/2016 12:09 AM, John Harshman wrote:
> I'm sorry, but this is incorrect in regard to species. It would be true
> if evolution were perfectly clocklike, b

no that is not true. time is not a measured characteristic in closest
path analysis. It might be that two species are a billion years
separated, but still closely related, and the shortest path analysis
will demonstrate that relationship.

[ruben safir]

On 04/14/2016 12:09 AM, John Harshman wrote:

> If lineages evolve at different rates, the two most similar species may
> not be the two most closely related. If you want to learn something
> about how phylogenetics is done, and why, I could recommend a book or two.

I've read and it is GOD AWFUL. Few fields can benefit from mathematical
analysis as much as paleontology. You guys make dozens of wrong
assumptions, like for example it matters that life is a rooted tree or
that time is a factor in determining relationships between objects.

[ruben safir]

On 04/11/2016 07:07 PM, John Harshman wrote:
> Dijkstra's algorithm appears to produce a Steiner tree,

wrong, BTW.

[ruben safir]

On 04/11/2016 07:07 PM, John Harshman wrote:

> That just doesn't bear much resemblance to the actual trees of
> relationships among species.

the method is valid for a tree, a square and tetrahydral... it doesn't
matter, or a factorial algorithm.

[ruben safir]

On 04/11/2016 07:35 PM, RSNorman wrote:
> You don't need Dijkstra's algorithm to discover the distance between
> any two objects on a simply connected rooted tree,

No, you can't determine where objects are attached to a tree without
some analysis like Dijkstra's algorithm.

[RSNorman]

On Thu, 14 Apr 2016 16:23:25 -0400, ruben safir <ru...@mrbrklyn.com>
wrote:

I am not sure you understand paleontology.

A basic concept of evolution is that any two species have a 'most
recent common ancestor'. The shortest evolutionary path between the
two is to back up the tree from each until you reach that common
ancestor. The path is then up one leg and down the other.

A simple way of finding it, although not the quickest, is to back up
from each species until you get the common ancestor of "all life" or,
more likely "all eukaryotes" or "all bilateral animals". The two
paths will necessarily have a long piece at the top that is shared by
both. So the shortest connecting path deletes the shared portion and
starts from where they diverge, the most recent common ancestor of
those two.

It doesn't matter if different tetrapods derive from diferent
predecessors. The only evolutionary path between then means going
back to the one common ancestor.

You may have a different concept of "shortest path" that involves
taking short cuts. Maybe you are thinking of going from a human to a
robin simply by taking a short cut across "any old tetrapod" because
humans and robins are both tetrapods (despite having two legs). No,
you have to go back to whatever the common ancestor was.

[Peter Nyikos]

Good point. The lineages started diverging soon enough; how can we be
sure the LCA was anything more than panderychthian fish, for example?

Tiktaalik may have belonged to one lineage, some yet undiscovered
transitional form of the same "grade" might have belonged to another.

> I'm not sure you really understand graph theory after reading the
> previous posts.
>
> Ruben

We are essentially in the realm of directed graphs whose branch points
form what set theorists call trees: partially ordered sets such that
the predecessors of every element are well-ordered. And since the
so-called "phylogenetic trees" have data only at the branch tips,
we can simply say that the elements below any branch tip are
well-ordered.

Did Norman say anything that goes against this?

Peter Nyikos
Professor, Dept. of Mathematics -- standard disclaimer--
University of South Carolina
Specialty: set-theoretic topology

PS It is great to have you back, Ruben! Is it just a coincidence that
you've reappeared after I replied today to your Feb. 8 post on
the Deccan Traps?

[John Harshman]

I'm not even sure what you're replying to there. What data points? What
path?

[John Harshman]

On 4/14/16 1:48 PM, Peter Nyikos wrote:
> On Thursday, April 14, 2016 at 4:23:15 PM UTC-4, ruben safir wrote:
>> On 04/14/2016 12:07 AM, RSNorman wrote:
>
>>>> Yeah you do, because it demonstrates the shortest path and in fact there
>>>>> ARE loops in the data and determining the shortest path is the only means
>>>>> of determining the trees shap.
>
>>> Perhaps I used the phrase "simply connected" from topology in a way
>>> that does not apply to graph theory.
>>>
>>> The biological tree of descent is necessarily acyclic. There are no
>>> loops or cycles in the data.
>>
>>
>> The result is, the data is NOT, which is why it is essential to find the
>> shortest path
>>
>> which is the essential problem.
>>
>> It might be that the resulting analysis is not either, because it is an
>> assumption that probably isn't true. There is no reason to believe that
>> a single tetrapod came out of the water to spawn all tetrapods. It
>> could very well be a group of closely relates tetrapods, that might even
>> have been geographically isolated from each other.
>
> Good point. The lineages started diverging soon enough; how can we be
> sure the LCA was anything more than panderychthian fish, for example?

Why, by doing phyogenetic analysis and estimating the character states
of that ancestor.

> Tiktaalik may have belonged to one lineage, some yet undiscovered
> transitional form of the same "grade" might have belonged to another.

To be sure, there were doubtless plenty of branchings near the root of
tetrapods that had various combinations of primitive and derived
characters, some of which we know about. But it seems pointless to posit
unknown fossils that would upset our current understanding. Doesn't mean
they don't exist, just that we have no reason to think about them.

[John Harshman]

There are many algorithms used in phylogenetic analysis. Dijkstra's is
not one of them. Are you arguing that it should be? If so, why?

[John Harshman]

No it won't, unless by "closely related" you just mean "similar". But
what the term means in phylogenetics is "sharing a more recent common
ancestry than with some third taxon", i.e. the sort of thing
phylogenetic trees say. There is no particular justification for making
trees that just measure similarity. (There are methods for making such
trees, e.g. UPGMA, but why would you use them?)

[John Harshman]

On 4/14/16 1:30 PM, ruben safir wrote:
> On 04/14/2016 12:09 AM, John Harshman wrote:
>> If lineages evolve at different rates, the two most similar species may
>> not be the two most closely related. If you want to learn something
>> about how phylogenetics is done, and why, I could recommend a book or two.
>
> I've read and it is GOD AWFUL.

What exactly have you read?

> Few fields can benefit from mathematical
> analysis as much as paleontology. You guys make dozens of wrong
> assumptions, like for example it matters that life is a rooted tree or
> that time is a factor in determining relationships between objects.
>

Life isn't a rooted tree? How so? And almost nobody thinks time is a
factor in determining relationships. There are methods that make use of
time, e.g. Stratocladistics, but they are seldom used.

I don't think you have any idea how phylogenetic analysis is done, or
how it ought to be done.

Try Inferring Phylogenies by Joe Felsenstein for some instruction. There
should be enough math there to satisfy you.

[John Harshman]

I don't think that's the case. For one thing, it seems not to allow for
nodes that aren't present in the initial data set. Is that true?

[John Harshman]

Agreed. I meant "minimum-spanning tree".

[RSNorman]

On Thu, 14 Apr 2016 16:34:43 -0400, ruben safir <ru...@mrbrklyn.com>
wrote:

My bad. I should not have said "any two objects". You have to
actually select the objects first. Perhaps in abstract computer
science you are given some descriptors for the two nodes and then have
to search the tree, first, to see where they are located. In actual
biology, you essentially point to the nodes on the tree and say 'how
did I get from here to there?'

[Peter Nyikos]

Cladistics isn't just cladistic taxononmy. Why wasn't this
obvious to you?

> What I don't know is that your main
> complaint is the absolute adherence to rank 1 classifications.

Your statement is true as it stands: you don't know this, because
I have no such complaint, never did. But that's not what you
meant to say, is it?

I gave facts about rank 1 systems in general, and you somehow think
that I had something against them *per se*. Where you get this idea,
given that "rank 1" describes BOTH systems which I would have
consistently wanted to be used and developed, is beyond me.

> But
> traditional classifications are rank 1 classifications. Nobody at all is
> proposing any other sort of classification. Thus I am confused.

Yes, you most certainly are, and I hope this post finally eliminates
the confusion that a little more attention to my OP would have
obviated.

> > Cal King was someone I'd love to have return to sci.bio.paleontology;
> > he wasn't much for off-topic stuff or personal attacks.
>
> Wasn't that just a bit of off-topic stuff?

Talking about s.b.p. participants in approving ways for their on-topic
contributions may not agree with a letter-of-the-law narrow interpretation
of "on topic," but I think it is very much OK, and besides it isn't
the kind of off-topic stuff I was talking about.

OTOH you made a very derogatory comment about Cal King
not long after I returned to s.b.p. at the end of 2010. Would
you like to tell us your current opinion of him?

Concluded in next reply.

Peter Nyikos
Professor, Dept. of Mathematics -- standard disclaimer--

Univ. of South Carolina at Columbia

[Peter Nyikos]

On Wednesday, April 13, 2016 at 3:18:27 PM UTC-4, John Harshman wrote:
> On 4/13/16 7:16 AM, Peter Nyikos wrote:
> > On Tuesday, April 12, 2016 at 9:55:54 AM UTC-4, John Harshman wrote:

Picking up where I left off in my first reply.


> >> What your main complaint seems to be is
> >> that the rank 1 classifications in cladistic classfication (as opposed
> >> to tree-building methods) aren't allowed to include paraphyletic groups.
> >
> > You are conflating two utterly different things here. It would defeat
> > the main practical application of the rank 1 cladistic classification --
> > reconstructing phlogeny -- to include paraphyletic groups.
>
> I think you mean "representing" rather than "reconstructing".

I meant reconstructing. Representing something without a guide
to connecting the representation with the thing itself is not
of much use.

> But of
> course. That's why we don't like paraphyletic groups.

Because they don't fit the only thing you care about,
which is finding trees that show the topological pattern in which
lineages diverge from each other.

> > That
> > is also the main reason why I am all for two parallel classification
> > schemes, and that is why you are completely off base with what my main
> > complaint "seems to be".
>
> See, it's very confusing that you complain about rank 1 classifications

I don't, and never did. I hope my first reply to this post of yours
made that clear, once and for all.


> but your proposal is to use two rank 1 classifications. It's like saying
> you hate apples, when what you mean is that you don't like having just
> one apple; you want both a pippin and a macintosh.

This is pure GIGO from beginning to end.

>
> >> Just what's wrong with rank 1 classifications is unclear, because you
> >> have [not] told us yet.
> >
> > Come off it. I have made it clear that the traditional system
> > is ALSO a rank 1 system, but was never intended for depicting
> > phylogeny without the aid of "trees of bubble taxa"-- something
> > you couldn't even recognize as trees, even though they depicted
> > phylogeny.
>
> Bubble diagrams aren't trees.

They are trees in the mathematical sense, with the "elements"
of which I wrote to Ruben less than an hour ago being
the taxa that the individual bubbles depict. The connecting
lines depict the individual lineages which go from one taxon
to the next, if they are solid. If they are dashed, they
simply show that the best available evidence is that an ancestor
of the taxon at the upper end of the dashed line can be found
in the taxon at the lower end.

<small snip>

> > My real "cladistic" beef is the prohibition by
> > your kind of paraphyletic taxa, except -- very reluctantly -- genera.
>
> What do you mean by "your kind of paraphyletic taxa"?

Wrong parsing. "your kind" prohibits paraphyletic taxa.

<small snip>

> > One disagreement between you and Cal King centered on his claim
> > that the king snake genus *Lampropeltis* is paraphyletic, and
> > that *Getulus* is a daughter taxon.
> >
> > I don't recall why you disagreed with him on that, and would
> > like to know why.
>
> I have no idea. In modern practice, by the way, we reject paraphetic
> genera as much as any paraphyletic taxon.

...as long as you are sure they are paraphyletic, which occurs
very rarely, as your erstwhile disagreement with Cal illustrates.

Peter Nyikos
Professor, Dept. of Mathematics -- standard disclaimer--

Univ. of South Carolina, Columbia
http://people.math.sc.edu/nyikos/

[John Harshman]

On 4/14/16 2:41 PM, Peter Nyikos wrote:
> On Wednesday, April 13, 2016 at 3:18:27 PM UTC-4, John Harshman wrote:
>> On 4/13/16 7:16 AM, Peter Nyikos wrote:
>>> On Tuesday, April 12, 2016 at 9:55:54 AM UTC-4, John Harshman wrote:
>
> Picking up where I left off in my first reply.
>
>
>>>> What your main complaint seems to be is
>>>> that the rank 1 classifications in cladistic classfication (as opposed
>>>> to tree-building methods) aren't allowed to include paraphyletic groups.
>>>
>>> You are conflating two utterly different things here. It would defeat
>>> the main practical application of the rank 1 cladistic classification --
>>> reconstructing phlogeny -- to include paraphyletic groups.
>>
>> I think you mean "representing" rather than "reconstructing".
>
> I meant reconstructing. Representing something without a guide
> to connecting the representation with the thing itself is not
> of much use.

Perhaps you mean something different by "cladistic classification" than
I do. You have to separate classification from tree-building. It's the
latter that reconstructs phylogeny and the former that represents
phylogeny (i.e. is abstracted from the reconstructed tree). You may not
mean to conflate the two, but that's exactly what your language is doing
here. Perhaps this is a feature of mixing mathematical terms ("rank 1
classification") with biological ones ("cladistic"). I would suggest
abandoning the whole "rank 1 classification" terminology as
counterproductive.

>> But of
>> course. That's why we don't like paraphyletic groups.
>
> Because they don't fit the only thing you care about,
> which is finding trees that show the topological pattern in which
> lineages diverge from each other.

Not the only thing I care about, just the only thing that trees are
intended to represent. And that we can't easily represent more than one
parameter clearly in a nested classification. I pick the one that is
most useful rather than trying to mush together two incompatible ones.

>>> That
>>> is also the main reason why I am all for two parallel classification
>>> schemes, and that is why you are completely off base with what my main
>>> complaint "seems to be".
>>
>> See, it's very confusing that you complain about rank 1 classifications
>
> I don't, and never did. I hope my first reply to this post of yours
> made that clear, once and for all.

No, it didn't. This is what confused me, and I'd like to know what you
meant:

"Anyway, a rank 1 classification system works perfectly for extant taxa,
but it gets worse and worse the further back one goes in time. For
instance, the only taxa containing *Panderichthys* are the genus itself
and a super-class that includes all terrestrial vertebrates.

I'll have more to say about the drawbacks of rank 1 classification
systems for paleontologists in later posts to this thread."

That sure sounds like you have a problem with rank 1 classifications
when there are extinct species involved.

>> but your proposal is to use two rank 1 classifications. It's like saying
>> you hate apples, when what you mean is that you don't like having just
>> one apple; you want both a pippin and a macintosh.
>
> This is pure GIGO from beginning to end.

It's an analogy intended to make my assertion clear to you. Why is my
assertion wrong?

>>>> Just what's wrong with rank 1 classifications is unclear, because you
>>>> have [not] told us yet.
>>>
>>> Come off it. I have made it clear that the traditional system
>>> is ALSO a rank 1 system, but was never intended for depicting
>>> phylogeny without the aid of "trees of bubble taxa"-- something
>>> you couldn't even recognize as trees, even though they depicted
>>> phylogeny.
>>
>> Bubble diagrams aren't trees.
>
> They are trees in the mathematical sense, with the "elements"
> of which I wrote to Ruben less than an hour ago being
> the taxa that the individual bubbles depict. The connecting
> lines depict the individual lineages which go from one taxon
> to the next, if they are solid. If they are dashed, they
> simply show that the best available evidence is that an ancestor
> of the taxon at the upper end of the dashed line can be found
> in the taxon at the lower end.

I don't know that they actually are trees. Is a bubble equivalent to a
line segment? I know they're intended to give an impression of
treeishness, at least, and to represent species diversity over time. But
they really aren't very useful as trees.

>>> My real "cladistic" beef is the prohibition by
>>> your kind of paraphyletic taxa, except -- very reluctantly -- genera.
>>
>> What do you mean by "your kind of paraphyletic taxa"?
>
> Wrong parsing. "your kind" prohibits paraphyletic taxa.

Oh. I see. By "your kind" you refer to almost all practicing
systematists, right?

>>> One disagreement between you and Cal King centered on his claim
>>> that the king snake genus *Lampropeltis* is paraphyletic, and
>>> that *Getulus* is a daughter taxon.
>>>
>>> I don't recall why you disagreed with him on that, and would
>>> like to know why.
>>
>> I have no idea. In modern practice, by the way, we reject paraphetic
>> genera as much as any paraphyletic taxon.
>
> ...as long as you are sure they are paraphyletic, which occurs
> very rarely, as your erstwhile disagreement with Cal illustrates.

No, it occurs quite frequently, more and more often as DNA sequencing
increases. Why, around half of all pre-sequencing bird genera* are
paraphyletic or polyphyletic. Of course this is in the process of being
fixed.

*I omit monotypic genera from this count, as they are incapable of being
paraphyletic.

[John Harshman]

Well of course it's obvious to me. But why did you bring it up? Please
try to make some kind of clear point, preferably one that has to do with
the subject. I will freely admit that you have heard the terms MP, NJ,
and PAUP. But why is that relevant here?

>> What I don't know is that your main
>> complaint is the absolute adherence to rank 1 classifications.
>
> Your statement is true as it stands: you don't know this, because
> I have no such complaint, never did. But that's not what you
> meant to say, is it?

> I gave facts about rank 1 systems in general, and you somehow think
> that I had something against them *per se*. Where you get this idea,
> given that "rank 1" describes BOTH systems which I would have
> consistently wanted to be used and developed, is beyond me.

It's beyond me too. But look above:

"Anyway, a rank 1 classification system works perfectly for extant taxa,
but it gets worse and worse the further back one goes in time. For
instance, the only taxa containing *Panderichthys* are the genus itself
and a super-class that includes all terrestrial vertebrates.

I'll have more to say about the drawbacks of rank 1 classification
systems for paleontologists in later posts to this thread."

This seems to me to be a complaint about applying rank 1 classifications
to paleontology. If that isn't what it is, what is it?

>> But
>> traditional classifications are rank 1 classifications. Nobody at all is
>> proposing any other sort of classification. Thus I am confused.
>
> Yes, you most certainly are, and I hope this post finally eliminates
> the confusion that a little more attention to my OP would have
> obviated.

No, it doesn't. All you tell me here is that I don't understand what
you're saying, but you don't explain what you meant.

>>> Cal King was someone I'd love to have return to sci.bio.paleontology;
>>> he wasn't much for off-topic stuff or personal attacks.
>>
>> Wasn't that just a bit of off-topic stuff?
>
> Talking about s.b.p. participants in approving ways for their on-topic
> contributions may not agree with a letter-of-the-law narrow interpretation
> of "on topic," but I think it is very much OK, and besides it isn't
> the kind of off-topic stuff I was talking about.
>
> OTOH you made a very derogatory comment about Cal King
> not long after I returned to s.b.p. at the end of 2010. Would
> you like to tell us your current opinion of him?

My current opinion is my previous opinion, having had no additional
information since then. But I really have no interest in discussing Cal
King. Why would anyone?

[Popping mad]

On Thu, 14 Apr 2016 13:55:46 -0700, John Harshman wrote:

>> irrelevant to the analysis. It doesn't make a difference if there is a
>> rooted tree or not. what matters is the shortest path between data
>> points.
>
> I'm not even sure what you're replying to there. What data points? What
> path?

Data points is observations made about a speicimen that can be measured
and described. The path is, in this case, a theoretical path between two
described object. Graphics analysis allgorthms do not determine the show
of the data, and saying they can't be used on a tree structure of
objects, is like saying you can't use addition or algrebra on a Tree
because it doesn't describe such shapes. its apples and oranges

[Popping mad]

On Thu, 14 Apr 2016 14:11:11 -0700, RSNorman wrote:

> n actual biology, you essentially point to the nodes on the tree and say
> 'how did I get from here to there?'

that fails us all the time. You should start with two species and say,
and we don't know where they are on the tree of life, so lets describe
them and many other similar objects and discover there relationships.
That discovery can be done with objective analysis though graphic
algorthims.

[Popping mad]

On Thu, 14 Apr 2016 14:03:14 -0700, John Harshman wrote:

> It might be that two species are a billion years
>> separated, but still closely related, and the shortest path analysis
>> will demonstrate that relationship.
>>
> No it won't, unless by "closely related" you just mean "similar".

Really? Horse Shoe crabs are more closely related to extinct Precambrian
trilibites or Homo Sapians?

Time is not a factor in such analysis

[Popping mad]

On Thu, 14 Apr 2016 14:07:32 -0700, John Harshman wrote:

>> wrong, BTW.
>>
>>
> Agreed. I meant "minimum-spanning tree".

no it has zero to do with trees, circles or boxes.

[Popping mad]

On Thu, 14 Apr 2016 14:07:10 -0700, John Harshman wrote:

> I don't think that's the case.

But your wrong

> For one thing, it seems not to allow for
> nodes that aren't present in the initial data set. Is that true?

IF what your saying is that it depends of a finite set that it is
analysing, then that is true. And when you add another object then you
need to do the analysis again, just like any of mathamtical function.

[ruben safir]

On 04/14/2016 04:43 PM, RSNorman wrote:
> A basic concept of evolution is that any two species have a 'most
> recent common ancestor'. The shortest evolutionary path between the
> two is to back up the tree from each until you reach that common
> ancestor. The path is then up one leg and down the other.
>
> A simple way of finding it, although not the quickest, is to back up
> from each species until you get the common ancestor of "all life" or,
> more likely "all eukaryotes" or "all bilateral animals".

no, you can analyze them correctly rather than depending on a
construction of tree which is highly speculative and inaccurate.

[ruben safir]

On 04/14/2016 04:48 PM, Peter Nyikos wrote:
> We are essentially in the realm of directed graphs whose branch points
> form what set theorists call trees: partially ordered sets such that
> the predecessors of every element are well-ordered. And since the
> so-called "phylogenetic trees" have data only at the branch tips,
> we can simply say that the elements below any branch tip are
> well-ordered.
>
> Did Norman say anything that goes against this?

that is only one way of analyzing them and it depends that the objects
are inserted correctly IN THE TREE. And that is the problem. They are
not inserted correctly in the tree and we don't even know the shape of
the tree.


What we do know, it have we can compare a group of related species, or
any group os species, and determine how closely they are related by
determining the shortest path between any two. Then you can try to fit
it in a tree.

Of course, you are the math professor and I am just a student.

We can say with a fair degree of certainty that species A is more
closely related to species B than species C is.

The direction and the shape of the overall relationship of all life on
this planet since the beginning of time is irrelevant to this analysis,
and that is why they keep following blind alleys.

Ruben

[RSNorman]

You seem to think the problem is something like:

Let's see, now. I have
something with wings, an exoskeleton, jointed legs
something with wings, an endoskeleton, jointed legs
something with no wings, no skeleton, no legs
something with no wings, no skeleton, legs
something with no wings, endoskeleton, no legs
No what kind of tree is that going to make?

That sort of thing (though rather more refined) is a phenetic analysis
which is now generally considered unsuitable for determining
phylogeny, true evolutionary relations expressed in a tree of common
descent.

In phenetics, lilzards and crocodilians are very similar (both being
reptiles) and both very different from birds in many aspects.
Unfortunately, all the evidence indicates that birds and crocodilians
are rather closely related evolutionarily with lizards being rather
distant.

The trees we are talking about here are those that reflect
evolutionary relationships, not mere similarity of features. The
problem in trying to figure out phylogeny from similarities is that
some features seemingly minor in appearance count heavily while others
indicating very striking change in appearance (morphology) are much
less important biologically..

[ruben safir]

On 04/14/2016 04:59 PM, John Harshman wrote:
> But it seems pointless to posit unknown fossils that would upset our
> current understanding.

We have no current understanding. We have a half dozen or so fossils
from billions of individuals over millions of years. That is part of
the problem, right there. It is arrogance to conclude our understanding
of the relationships of known tetrapods and later species beyond the
limits of what these exact fossils tell us, which is painfully little, I
might add.

>Doesn't mean they don't exist, just that we have
> no reason to think about them.

Not for a thoughtful scientist who understands how small his sample is.

[ruben safir]

On 04/14/2016 05:00 PM, John Harshman wrote:
> Dijkstra's is not one of them. Are you arguing that it should be? If so,
> why?

Ah - now that is a good question. From what I've seen it is better than
the currently applied methods because of its focus. But there are
likely better algorithms currently in existence, and a good application
of the theory to the problem will likely create better ones.

But the reason fir its superiority of what I've seen is its lack of
presumptions.

[ruben safir]

On 04/14/2016 05:03 PM, John Harshman wrote:
>>
> No it won't, unless by "closely related" you just mean "similar".

so closely related is not similar?

[RSNorman]

On Thu, 14 Apr 2016 22:53:22 -0400, ruben safir <ru...@mrbrklyn.com>
wrote:

recent common ancestor'. Do you disagree?

Is the evolutionary tree of ancestry and descent not necessarily a
directed acyclic (except for some very unusual exceptions) tree?

Some of our trees may be somewhat speculative and inaccurate but there
are techniques to produce the trees and to evaluate the relative
validity of one tree over another.

Perhaps you do not understand that the evolutionary (phylogenetic)
tree is supposed to reflect biology the way that biology works, not
the way that mathematics aggregates sets.

[ruben safir]

On 04/14/2016 11:31 PM, RSNorman wrote:
> A basic concept of evolution is that any two species have a 'most

> recent common ancestor'. Do you disagree?

for the analysis of the relation between species this doesn't matter.
You define their relationship according to their shortest path and how
that compares to other specimens.

I can't say this any clearer, or at least I can't think of any way. You
keep throwing irrelevant facts into the discussion. We know they are
closest in relationship because they have the highest number of similar
traits. so the distance between them is short Now if you have ten
species that are closely related, you should likewise be able to tell
which are more closely related to others, and if you want to, to
eventually build a tree. The tree structure between any species is
conjecture. the relationships though are measurable, quantifiable, can
be determined with standard graph theory.


Also, if the closest path between two species goes through a third
species, your hierarchy is now viable and that is one of the major
advantages of the shortest path theory and doctrine, and why it is
considered revolutionary, even still now 40+ years after the its
publication.

[ruben safir]

On 04/14/2016 04:55 PM, John Harshman wrote:
> What data points? What path?

data points in this case are species and the path is edges between two
data points, that has a length or cost defined as say, the number of
evolutionary differences between the two species. Data points,
confusingly are also be refereed by me as the collection of facts about
any set species, that can separate them one from another. I suppose I
was a little bit confusing there.

sorry

[ruben safir]

On 04/14/2016 11:05 PM, RSNorman wrote:
> No what kind of tree is that going to make?

you almost have it but when you ask what kind of tree that makes, I sat
back in my chair and laughed. I'm sorry RS. You can bring a horse to
water, but you can't make them drink.

[ruben safir]

On 04/14/2016 11:05 PM, RSNorman wrote:

> The trees we are talking about here are those that reflect
> evolutionary relationships

You don't have a clue what that looks like and promoting it as evidence
is bad science. You do, however, have the physical evidence before you
which which can be demonstrated to have known relationships,
mathematically provable, every bit as much as 1 and 1 is 2.

You think you know everything and your not listening. If the shortest
path shows that the relationship between species doesn't agree with your
"tree" then the three is WRONG, and the Shortest Path algorithm is
correct. It can never happen in the reverse.

[John Harshman]

Here's where your difficulty writing in English obscures what you are
trying to say. I don't know what "show of the data" means. If data
points are individual observations rather than collections of
observations, I don't see why the distance between them should concern
us at all. Perhaps you should pull back and explain why what you're
proposing is superior to what is commonly being done in paleontology
right now. To do that you have to be acquainted with what is done now.
Are you?

[John Harshman]

Let me ask one simple question: are you proposing to connect species to
species without any intervening, internal nodes?

[John Harshman]

It seems here that all you're doing is calculating pairwise distances.
And you are redefining "relationships" to be something entirely
non-biological. Don't expect to interest any biologists in that, or even
any mathematicians interested in biology.

[John Harshman]

I don't believe you know anything about currently applied methods.

[John Harshman]

That's correct. What you are proposing is a redefinition of the term
"related" that has nothing to do with biology. There was in fact a
movement of that sort in systematics in the late 1960s and early 1970s,
called "phenetics". Pheneticists too decided that phylogeny was
unknowable and decided to classify purely on the basis of overall
similarity. That movement died long ago for a number of reasons. First,
overall similarity is a subjective concept that can vary radically
depending on what data you choose. Second, estimating phylogeny is not
in fact hopeless.

Again I advise you to read Felsenstein's book.

[John Harshman]

Then nobody should care, as it's trees we are interested in.

[John Harshman]

No, that isn't at all what I'm saying. In phylogenetics, the actual data
is at the tips of the trees, and internal nodes are hypothetical,
inferred from the data. If your methods allow no such inferences, your
methods won't work to discover relationships among species. And note
that "relationships" refers to actual order of descent, a biological
property rather than an arbitrary mathematical one.

[RSNorman]

On Fri, 15 Apr 2016 02:46:14 -0400, ruben safir <ru...@mrbrklyn.com>
wrote:

I write " A basic concept of evolution is ..."
You write "for the analysis of the relation between species this
doesn't matter"

"Consider a spherical cow". Look it up if you don't understand what
I mean.

[RSNorman]

On Fri, 15 Apr 2016 03:01:56 -0400, ruben safir <ru...@mrbrklyn.com>
wrote:

An examination of the molecular genetics between species has many
times shown that comparing things "because they are similar" was very
wrong.

Bacteria and fungi and plants have cell walls; animals do not. So
make two categories.

Earthworms and centipedes are segmented, molluscs are not. So
molluscs must be in a different category from the other two.

Nematodes and rotifers are pseudocoelomates with fixed cell count and
completely determined development. They must be lumped together.

I already gave Popping mad the example of phyletics where crocs and
lizards are put together because of a very large number of
similarities that make them both very different from birds.

Every one of the examples I list above was at one time the way
biologists evaluated relationships and drew trees. We change all the
time. But the only important thing is "what tree best explains the
evolutionary relationship." That relationship is ancestry or, if you
prefer, descent (with modification). Nothing else matters.

Ignoring the biology means you do not produce results useful to
biology.

[Popping mad]

On Fri, 15 Apr 2016 07:15:42 -0700, John Harshman wrote:

> That's correct. What you are proposing is a redefinition of the term
> "related" that has nothing to do with biology. There was in fact a
> movement of that sort in systematics in the late 1960s and early 1970s,
> called "phenetics". Pheneticists too decided that phylogeny was
> unknowable and decided to classify purely on the basis of overall
> similarity. That movement died long ago for a number of reasons. First,
> overall similarity is a subjective concept that can vary radically
> depending on what data you choose. Second, estimating phylogeny is not
> in fact hopeless.

that just means they didn't do it correctly. The suggestion made
previously, for example about Lizzards, Birds and Crocs is laughable.

Aside which this is not similary scores, but the shortest path. You
can't take a path, for example, between a Pakicetids and a shark and a
whale without going THROUGH the auditory bulla anatomy no matter how
many similarities there are between whales and sharks.

[Popping mad]

On Fri, 15 Apr 2016 08:38:36 -0700, RSNorman wrote:

>>You think you know everything and your not listening. If the shortest
>>path shows that the relationship between species doesn't agree with your
>>"tree" then the three is WRONG, and the Shortest Path algorithm is
>>correct. It can never happen in the reverse.
>
> An examination of the molecular genetics between species has many times
> shown that comparing things "because they are similar" was very wrong.

No that is not what it showed. What if showed was the examination of
data was wholey inadequate until genetic comparisions came along, which
in retrospect, should have been obvious. That is exactly an example
where the application of the shortest path methodology would have
prevented a huge error. generic mapping is not OUTSIDE of the data
analysis, but an essential part of it.

[John Harshman]

You must understand that the shortest path between two points has little
to do with the phylogenetic relationships of two species. Do you know
what a patristic distance is? And the two most similar species may not
be the two most closely related. This is as true for molecular data as
for morphological data. An emu, for example, is more similar to an
ostrich than to a tinamou in pretty much any genetic comparison you care
to make. And yet the emu and the tinamou are more closely related than
either is to the ostrich. That's because tinamous have an unusually
rapid rate of molecular evolution.

[John Harshman]

On 4/15/16 11:25 AM, Popping mad wrote:
> On Fri, 15 Apr 2016 07:15:42 -0700, John Harshman wrote:
>
>> That's correct. What you are proposing is a redefinition of the term
>> "related" that has nothing to do with biology. There was in fact a
>> movement of that sort in systematics in the late 1960s and early 1970s,
>> called "phenetics". Pheneticists too decided that phylogeny was
>> unknowable and decided to classify purely on the basis of overall
>> similarity. That movement died long ago for a number of reasons. First,
>> overall similarity is a subjective concept that can vary radically
>> depending on what data you choose. Second, estimating phylogeny is not
>> in fact hopeless.
>
> that just means they didn't do it correctly. The suggestion made
> previously, for example about Lizzards, Birds and Crocs is laughable.

I presume it was intended as an analogy only, not a serious suggestion.
To my knowledge nobody has proposed a phylogeny in which birds and crocs
are not both archosaurs. Nor do I recall any pheneticist doing an
analysis of reptiles. One can however point to real-world examples in
which overall similarity failed as a method. I have mentioned in another
post the case of paleognath birds, in which unusually rapid evolution in
one group (tinamous) led to their being misplaced in many analyses, just
because the shortest path wasn't the true index of closeness of
relationships.

> Aside which this is not similary scores, but the shortest path. You
> can't take a path, for example, between a Pakicetids and a shark and a
> whale without going THROUGH the auditory bulla anatomy no matter how
> many similarities there are between whales and sharks.
>

I'm not even sure what sort of path you have in mind. Again, I urge you
to compare what you propose, whatever that is, with what systematists do
now and let us know what's different and why yours is better. To do that
you will first have to learn what systematists do now.

[ruben safir]

On 04/15/2016 10:05 AM, John Harshman wrote:
> are you proposing to connect species to species without any intervening,
> internal nodes?

there might be other species in between, but yes. Species are nodes and
between species are edges.

[Peter Nyikos]

On Friday, April 15, 2016 at 10:18:46 AM UTC-4, John Harshman wrote:
> On 4/14/16 7:30 PM, Popping mad wrote:
> > On Thu, 14 Apr 2016 14:07:10 -0700, John Harshman wrote:
> >
> >> I don't think that's the case.
> >
> > But your wrong
> >
> >> For one thing, it seems not to allow for
> >> nodes that aren't present in the initial data set. Is that true?
> >
> > IF what your saying is that it depends of a finite set that it is
> > analysing, then that is true. And when you add another object then you
> > need to do the analysis again, just like any of mathamtical function.
> >
> No, that isn't at all what I'm saying. In phylogenetics, the actual data
> is at the tips of the trees,

The data on characters is at least partly to be found at the
nodes of a certain kind of cladogram. You told me this yourself
when I was under the impression that cladograms only had data
at the tips of the trees.

> and internal nodes are hypothetical,
> inferred from the data.

The usual claim is that they represent the Last Common Ancestors (LCAs)
of the clade that they are the initial node of.

Of course, you cannot be sure that the characters will all agree
with that of the real LCA. For instance, if you allow extinct taxa
in the tree, the best you can usually hope for is information about
the skeleton of the LCA.

And if one were to find a complete skeleton of an animal, and it
perfectly matched that deduced for the LCA, you would still be
dead set against saying that this skeleton is a "prime ancestor
candidate," wouldn't you?

Even the acknowledgement that the word "candidate" makes no
claims as to certainty doesn't mollify you.

Are you SURE most paleontologists would agree with you on
that? Or would silence in the wake of your dogmatic
pronouncement be interpreted as agreement?

Don't forget, paleontologists are very much dependent
on external, peer-reviewed funding, so that if enough
highly regarded (like Prum) cladophiles publicly voice your
opinion, such questions need to be asked in an atmosphere
of strict confidence as to answers.

> If your methods allow no such inferences, your
> methods won't work to discover relationships among species.

And even if they do allow them, what assurance is there that the
inferred relationships are correct?

> And note
> that "relationships" refers to actual order of descent, a biological
> property rather than an arbitrary mathematical one.

Indeed, matching of all bones to within a millimeter of a suitable
scale model of the hypothetical LCA is an arbitrary mathematical standard,
not a biological one, right?

Peter Nyikos
Professor, Dept. of Mathematics -- standard disclaimer--
Univ. of South Carolina, Columbia, SC
http://people.math.sc.edu/nyikos/

[John Harshman]

Might I suggest that this is biologically and paleontologically
incorrect? We don't know about the great majority of species that have
ever existed and have no fossils of them. This is even more obvious if
we're trying to find the relationships of extant species. They are not
connected through chains of fossils. They're connected through unknown,
extinct ancestors. That is, there are nodes for which we have no data,
and any method attempting to find relationships must take that into
account. Every ancestral species is such a node.

[John Harshman]

On 4/15/16 4:25 PM, Peter Nyikos wrote:
> On Friday, April 15, 2016 at 10:18:46 AM UTC-4, John Harshman wrote:
>> On 4/14/16 7:30 PM, Popping mad wrote:
>>> On Thu, 14 Apr 2016 14:07:10 -0700, John Harshman wrote:
>>>
>>>> I don't think that's the case.
>>>
>>> But your wrong
>>>
>>>> For one thing, it seems not to allow for
>>>> nodes that aren't present in the initial data set. Is that true?
>>>
>>> IF what your saying is that it depends of a finite set that it is
>>> analysing, then that is true. And when you add another object then you
>>> need to do the analysis again, just like any of mathamtical function.
>>>
>> No, that isn't at all what I'm saying. In phylogenetics, the actual data
>> is at the tips of the trees,
>
> The data on characters is at least partly to be found at the
> nodes of a certain kind of cladogram. You told me this yourself
> when I was under the impression that cladograms only had data
> at the tips of the trees.

No, that's not data. It's an estimate based on the data. The actual data
would be observations of the character states of real individuals.

>> and internal nodes are hypothetical,
>> inferred from the data.
>
> The usual claim is that they represent the Last Common Ancestors (LCAs)
> of the clade that they are the initial node of.

They are estimates of the states possessed by those ancestors.

> Of course, you cannot be sure that the characters will all agree
> with that of the real LCA. For instance, if you allow extinct taxa
> in the tree, the best you can usually hope for is information about
> the skeleton of the LCA.
>
> And if one were to find a complete skeleton of an animal, and it
> perfectly matched that deduced for the LCA, you would still be
> dead set against saying that this skeleton is a "prime ancestor
> candidate," wouldn't you?

Yes. Because it's a useless thing to say.

> Even the acknowledgement that the word "candidate" makes no
> claims as to certainty doesn't mollify you.

Correct.

> Are you SURE most paleontologists would agree with you on
> that? Or would silence in the wake of your dogmatic
> pronouncement be interpreted as agreement?

Pretty sure. I know a couple who might agree with you to a degree: John
Alroy and Nick Matzke. You can ask a paleontologist, if you know any.

> Don't forget, paleontologists are very much dependent
> on external, peer-reviewed funding, so that if enough
> highly regarded (like Prum) cladophiles publicly voice your
> opinion, such questions need to be asked in an atmosphere
> of strict confidence as to answers.

You are building a power structure from your imagination and nothing
else. There is no systematics inquisition.

>> If your methods allow no such inferences, your
>> methods won't work to discover relationships among species.
>
> And even if they do allow them, what assurance is there that the
> inferred relationships are correct?

One would suspect that the inferred relationships would be nearly
guaranteed not to be correct, if I understand at all what he's saying.

>> And note
>> that "relationships" refers to actual order of descent, a biological
>> property rather than an arbitrary mathematical one.
>
> Indeed, matching of all bones to within a millimeter of a suitable
> scale model of the hypothetical LCA is an arbitrary mathematical standard,
> not a biological one, right?

This is just something you made up, so there seems no reason to address
such an odd hypothetical. You also seem to be changing the subject from
Poppin Mad/ruben's* obsession to yours.

*You see that they're the same person, right?

[Peter Nyikos]

On Friday, April 15, 2016 at 7:27:14 PM UTC-4, John Harshman wrote:
> On 4/15/16 4:08 PM, ruben safir wrote:
> > On 04/15/2016 10:05 AM, John Harshman wrote:
> >> are you proposing to connect species to species without any intervening,
> >> internal nodes?
> >
> > there might be other species in between, but yes. Species are nodes and
> > between species are edges.
> >
> Might I suggest that this is biologically and paleontologically
> incorrect?

You may suggest it, but you would only be right if you
confined yourself to saying that this is incorrect
according to the reigning cladistic ideoology.

> We don't know about the great majority of species that have
> ever existed and have no fossils of them.

So? If we have a fairly complete sequence of fossils, like
the species-to-species linkages in some lines of the tree
of Equidae, then one is perfectly justified in doing what
Ruben said. Kathleen Hunt could have gone a good bit further than
she did in filling out the tree in her article in the
Talk.Origins Archive, but she restricted herself to
depicting genera at the nodes of her tree.

http://www.talkorigins.org/faqs/horses/horse_evol.html

If you read her article, you can see that she could
have listed a number of species at various nodes.

>This is even more obvious if
> we're trying to find the relationships of extant species. They are not
> connected through chains of fossils. They're connected through unknown,
> extinct ancestors.

Yes, like horses and tapirs. It would be interesting to know whether
the tree of Tapiroidea is as good as that of Equioidea.

And Plato would agree with you that the extinct ones making the
direct connection are "unknown" but would add that Kathleen
Hunt probably had the "right opinion" about the ones connecting
a species of Hyracotherium to Equus.

Reference: Plato's Socratic dialogue "Meno."

Hunt's sequence has:

Hyracotherium/Eohippus - Orohippus - Epihippus - Mesohippus -
Miohippus - Parahippus - Merychippus - Dinohippus - Plesippus - Equus.

One possible improvement is to put Kalobatippus between
Parahippus and Merychippus, but its fossils were too fragmentary
at the time of writing to be reasonably confident.

> That is, there are nodes for which we have no data,
> and any method attempting to find relationships must take that into
> account. Every ancestral species is such a node.

That last sentence is ideology, not biology.

Peter Nyikos
Professor, Dept. of Mathematics -- standard disclaimer--

Univ. of South Carolina at Columbia, SC
http://people.math.sc.edu/nyikos/

[John Harshman]

On 4/15/16 4:54 PM, Peter Nyikos wrote:
> On Friday, April 15, 2016 at 7:27:14 PM UTC-4, John Harshman wrote:
>> On 4/15/16 4:08 PM, ruben safir wrote:
>>> On 04/15/2016 10:05 AM, John Harshman wrote:
>>>> are you proposing to connect species to species without any intervening,
>>>> internal nodes?
>>>
>>> there might be other species in between, but yes. Species are nodes and
>>> between species are edges.
>>>
>> Might I suggest that this is biologically and paleontologically
>> incorrect?
>
> You may suggest it, but you would only be right if you
> confined yourself to saying that this is incorrect
> according to the reigning cladistic ideoology.

Sorry, but I don't think you have any idea what ruben/Popping is
proposing here or what systematists actually do instead. This is not a
subject on which you are informed.

>> We don't know about the great majority of species that have
>> ever existed and have no fossils of them.
>
> So? If we have a fairly complete sequence of fossils, like
> the species-to-species linkages in some lines of the tree
> of Equidae, then one is perfectly justified in doing what
> Ruben said. Kathleen Hunt could have gone a good bit further than
> she did in filling out the tree in her article in the
> Talk.Origins Archive, but she restricted herself to
> depicting genera at the nodes of her tree.
>
> http://www.talkorigins.org/faqs/horses/horse_evol.html
>
> If you read her article, you can see that she could
> have listed a number of species at various nodes.

She could have, but would she be right? Should you assume that you have
sampled all the relevant nodes? How is that good practice?

>> This is even more obvious if
>> we're trying to find the relationships of extant species. They are not
>> connected through chains of fossils. They're connected through unknown,
>> extinct ancestors.
>
> Yes, like horses and tapirs. It would be interesting to know whether
> the tree of Tapiroidea is as good as that of Equioidea.
>
> And Plato would agree with you that the extinct ones making the
> direct connection are "unknown" but would add that Kathleen
> Hunt probably had the "right opinion" about the ones connecting
> a species of Hyracotherium to Equus.

Is there any reason to recruit Plato into this discussion? If so, you
must explain the relevance.

> Reference: Plato's Socratic dialogue "Meno."
>
> Hunt's sequence has:
>
> Hyracotherium/Eohippus - Orohippus - Epihippus - Mesohippus -
> Miohippus - Parahippus - Merychippus - Dinohippus - Plesippus - Equus.
>
> One possible improvement is to put Kalobatippus between
> Parahippus and Merychippus, but its fossils were too fragmentary
> at the time of writing to be reasonably confident.
>
>> That is, there are nodes for which we have no data,
>> and any method attempting to find relationships must take that into
>> account. Every ancestral species is such a node.
>
> That last sentence is ideology, not biology.

I think you need to stop with the pejorative terms if you want to keep
this discussion civil. I'm willing to modify that to "The overwhelming
number of ancestral species are such nodes and we have no reliable way
to determine which ones are not".

[Peter Nyikos]

On Friday, April 15, 2016 at 7:35:58 PM UTC-4, John Harshman wrote:
> On 4/15/16 4:25 PM, Peter Nyikos wrote:
> > On Friday, April 15, 2016 at 10:18:46 AM UTC-4, John Harshman wrote:
> >> On 4/14/16 7:30 PM, Popping mad wrote:
> >>> On Thu, 14 Apr 2016 14:07:10 -0700, John Harshman wrote:
> >>>
> >>>> I don't think that's the case.
> >>>
> >>> But your wrong
> >>>
> >>>> For one thing, it seems not to allow for
> >>>> nodes that aren't present in the initial data set. Is that true?
> >>>
> >>> IF what your saying is that it depends of a finite set that it is
> >>> analysing, then that is true. And when you add another object then you
> >>> need to do the analysis again, just like any of mathamtical function.
> >>>
> >> No, that isn't at all what I'm saying. In phylogenetics, the actual data
> >> is at the tips of the trees,
> >
> > The data on characters is at least partly to be found at the
> > nodes of a certain kind of cladogram. You told me this yourself
> > when I was under the impression that cladograms only had data
> > at the tips of the trees.
>
> No, that's not data. It's an estimate based on the data. The actual data
> would be observations of the character states of real individuals.

Ah. Like observations of the bones of a "prime candidate ancestor"
by students and researchers, as opposed to mere "estimates based
on the data" for the characters of LCAs that they would get
from cladistic analyses.

Thanks for expressing a point I've been trying to make for years.

> >> and internal nodes are hypothetical,
> >> inferred from the data.
> >
> > The usual claim is that they represent the Last Common Ancestors (LCAs)
> > of the clade that they are the initial node of.
>
> They are estimates of the states possessed by those ancestors.

Thanks for the precise wording.

> > Of course, you cannot be sure that the characters will all agree
> > with that of the real LCA. For instance, if you allow extinct taxa
> > in the tree, the best you can usually hope for is information about
> > the skeleton of the LCA.
> >
> > And if one were to find a complete skeleton of an animal, and it
> > perfectly matched that deduced for the LCA, you would still be
> > dead set against saying that this skeleton is a "prime ancestor
> > candidate," wouldn't you?
>
> Yes. Because it's a useless thing to say.

Useless for YOU. But now read my first paragraph.

You are a systematist, not a paleontologist. Don't presume to
say what is useful to paleontologists and what is not.

> > Even the acknowledgement that the word "candidate" makes no
> > claims as to certainty doesn't mollify you.
>
> Correct.
>
> > Are you SURE most paleontologists would agree with you on
> > that? Or would silence in the wake of your dogmatic
> > pronouncement be interpreted as agreement?
>
> Pretty sure. I know a couple who might agree with you to a degree: John
> Alroy and Nick Matzke.

What about Benton? [I used to keep getting Denton and Benton confused;
I hope now I got it straight.]

> You can ask a paleontologist, if you know any.

I have corresponded with a few, but I'll be trying to add
those two authors, Deborah L. Rook and John P. Hunter,
who had the audacity to state the following in their
abstract of a peer-reviewed paper:

Our studies suggest that taeniodonts arose from Cimolestes
through Alveugena,

Reference:
"Rooting Around the Eutherian Family Tree: the Origin and Relations of
the Taeniodonta," Journal of Mammalian Evolution, March 2014,
Volume 21, Issue 1, pp 75-91
http://link.springer.com/article/10.1007/s10914-013-9230-9

> > Don't forget, paleontologists are very much dependent
> > on external, peer-reviewed funding, so that if enough
> > highly regarded (like Prum) cladophiles publicly voice your
> > opinion, such questions need to be asked in an atmosphere
> > of strict confidence as to answers.
>
> You are building a power structure from your imagination and nothing
> else. There is no systematics inquisition.

You have no idea how easy it is to rationalize acceptances
and rejections until you've attended some hiring debates.

Do you think Feduccia would have a ghost of a chance of
getting a grant if Prum were one of the referees?

> >> If your methods allow no such inferences, your
> >> methods won't work to discover relationships among species.
> >
> > And even if they do allow them, what assurance is there that the
> > inferred relationships are correct?
>
> One would suspect that the inferred relationships would be nearly
> guaranteed not to be correct, if I understand at all what he's saying.
>
> >> And note
> >> that "relationships" refers to actual order of descent, a biological
> >> property rather than an arbitrary mathematical one.
> >
> > Indeed, matching of all bones to within a millimeter of a suitable
> > scale model of the hypothetical LCA is an arbitrary mathematical standard,
> > not a biological one, right?
>
> This is just something you made up,

Strange, this absence of a Yes or No answer.

> so there seems no reason to address
> such an odd hypothetical.

Have you ever measured the bones of fossil equids?

> You also seem to be changing the subject from
> Poppin Mad/ruben's* obsession to yours.

And yours as well, obviously.

> *You see that they're the same person, right?

I never bothered to check the headers of their posts that closely.

Peter Nyikos
Professor of Mathematics
U. of S. Carolina, Columbia -- standard disclaimer--
http://people.math.sc.edu/nyikos/

[Popping mad]

On Fri, 15 Apr 2016 11:41:18 -0700, John Harshman wrote:

> just because the shortest path wasn't the true index of closeness of
> relationships.

Yes it is. It always is.

[John Harshman]

It seems as if you think you've scored a point, some kind of "gotcha"
moment, but for the life of me I can't think what it might be.

>>>> and internal nodes are hypothetical,
>>>> inferred from the data.
>>>
>>> The usual claim is that they represent the Last Common Ancestors (LCAs)
>>> of the clade that they are the initial node of.
>>
>> They are estimates of the states possessed by those ancestors.
>
> Thanks for the precise wording.
>
>>> Of course, you cannot be sure that the characters will all agree
>>> with that of the real LCA. For instance, if you allow extinct taxa
>>> in the tree, the best you can usually hope for is information about
>>> the skeleton of the LCA.
>>>
>>> And if one were to find a complete skeleton of an animal, and it
>>> perfectly matched that deduced for the LCA, you would still be
>>> dead set against saying that this skeleton is a "prime ancestor
>>> candidate," wouldn't you?
>>
>> Yes. Because it's a useless thing to say.
>
> Useless for YOU. But now read my first paragraph.

Your first paragraph communicates nothing to me that suggests "prime
ancestor candidate" is other than a useless term. Perhaps if you
explained just why it's useful rather than expecting me to see the light
by sheer power of your inspiration?

> You are a systematist, not a paleontologist. Don't presume to
> say what is useful to paleontologists and what is not.

But you're presuming to say what's useful to paleontologists. In what
way are you at once able to say I can't while you can? I'm not a
paleontologist, but I do read the paleo literature, I've taken quite a
few courses in paleontology, and I know quite a few paleontologists and
have talked to them on many relevant subjects.

You, on the other hand, have read a book or too by Romer and, perhaps,
Colbert.

>>> Even the acknowledgement that the word "candidate" makes no
>>> claims as to certainty doesn't mollify you.
>>
>> Correct.
>>
>>> Are you SURE most paleontologists would agree with you on
>>> that? Or would silence in the wake of your dogmatic
>>> pronouncement be interpreted as agreement?
>>
>> Pretty sure. I know a couple who might agree with you to a degree: John
>> Alroy and Nick Matzke.
>
> What about Benton? [I used to keep getting Denton and Benton confused;
> I hope now I got it straight.]

As far as I know, Benton doesn't try to identify ancestors. Don't know
him personally, though.

>> You can ask a paleontologist, if you know any.
>
> I have corresponded with a few, but I'll be trying to add
> those two authors, Deborah L. Rook and John P. Hunter,
> who had the audacity to state the following in their
> abstract of a peer-reviewed paper:
>
> Our studies suggest that taeniodonts arose from Cimolestes
> through Alveugena,
>
> Reference:
> "Rooting Around the Eutherian Family Tree: the Origin and Relations of
> the Taeniodonta," Journal of Mammalian Evolution, March 2014,
> Volume 21, Issue 1, pp 75-91
> http://link.springer.com/article/10.1007/s10914-013-9230-9

Let me know if you ever get one to say what you want.

>>> Don't forget, paleontologists are very much dependent
>>> on external, peer-reviewed funding, so that if enough
>>> highly regarded (like Prum) cladophiles publicly voice your
>>> opinion, such questions need to be asked in an atmosphere
>>> of strict confidence as to answers.
>>
>> You are building a power structure from your imagination and nothing
>> else. There is no systematics inquisition.
>
> You have no idea how easy it is to rationalize acceptances
> and rejections until you've attended some hiring debates.
>
> Do you think Feduccia would have a ghost of a chance of
> getting a grant if Prum were one of the referees?

Depends on the nature of the proposal. And anyway, the other panel
members might have something to say about it too. You seem to be
imagining that Rick Prum is some kind of paleontology god. He's just a guy.

>>>> If your methods allow no such inferences, your
>>>> methods won't work to discover relationships among species.
>>>
>>> And even if they do allow them, what assurance is there that the
>>> inferred relationships are correct?
>>
>> One would suspect that the inferred relationships would be nearly
>> guaranteed not to be correct, if I understand at all what he's saying.
>>
>>>> And note
>>>> that "relationships" refers to actual order of descent, a biological
>>>> property rather than an arbitrary mathematical one.
>>>
>>> Indeed, matching of all bones to within a millimeter of a suitable
>>> scale model of the hypothetical LCA is an arbitrary mathematical standard,
>>> not a biological one, right?
>>
>> This is just something you made up,
>
> Strange, this absence of a Yes or No answer.

Not so strange. Which is better, purple or margarine?

>> so there seems no reason to address
>> such an odd hypothetical.
>
> Have you ever measured the bones of fossil equids?

Not once. Why?

>> You also seem to be changing the subject from
>> Poppin Mad/ruben's* obsession to yours.
>
> And yours as well, obviously.

?

>> *You see that they're the same person, right?
>
> I never bothered to check the headers of their posts that closely.

It isn't the headers. It's what they say.

[John Harshman]

Despite my counterexamples?

[Popping mad]

On Fri, 15 Apr 2016 07:09:18 -0700, John Harshman wrote:

> It seems here that all you're doing is calculating pairwise distances.
> And you are redefining "relationships" to be something entirely
> non-biological.

that is part of it. The Allgorthms are only effective to demonstrate
shortest paths between data and edges. How you measure the distances or
costs is a paleontology question, The key is to give a series of species
quantifiable numbers that can be compared

[John Harshman]

How is this quantifiable number derived and what is that number attached
to? One species? A pair of species? And whatever is a path between data
and an edge?

[Peter Nyikos]

On Friday, April 15, 2016 at 8:00:32 PM UTC-4, John Harshman wrote:
> On 4/15/16 4:54 PM, Peter Nyikos wrote:
> > On Friday, April 15, 2016 at 7:27:14 PM UTC-4, John Harshman wrote:
> >> On 4/15/16 4:08 PM, ruben safir wrote:
> >>> On 04/15/2016 10:05 AM, John Harshman wrote:
> >>>> are you proposing to connect species to species without any intervening,
> >>>> internal nodes?
> >>>
> >>> there might be other species in between, but yes. Species are nodes and
> >>> between species are edges.
> >>>
> >> Might I suggest that this is biologically and paleontologically
> >> incorrect?
> >
> > You may suggest it, but you would only be right if you
> > confined yourself to saying that this is incorrect
> > according to the reigning cladistic ideoology.
>
> Sorry, but I don't think you have any idea what ruben/Popping is
> proposing here

Tell me how the context alters the plainly evident meaning of
the words he wrote, which seems to talk about what Kathleen
Hunt did, only with genera instead of species.

<snip insulting opinions>

> >> We don't know about the great majority of species that have
> >> ever existed and have no fossils of them.
> >
> > So? If we have a fairly complete sequence of fossils, like
> > the species-to-species linkages in some lines of the tree
> > of Equidae, then one is perfectly justified in doing what
> > Ruben said. Kathleen Hunt could have gone a good bit further than
> > she did in filling out the tree in her article in the
> > Talk.Origins Archive, but she restricted herself to
> > depicting genera at the nodes of her tree.
> >
> > http://www.talkorigins.org/faqs/horses/horse_evol.html
> >
> > If you read her article, you can see that she could
> > have listed a number of species at various nodes.
>
> She could have, but would she be right?

Are you sure she would be wrong in even ONE instance?

>Should you assume that you have
> sampled all the relevant nodes? How is that good practice?
>
> >> This is even more obvious if
> >> we're trying to find the relationships of extant species. They are not
> >> connected through chains of fossils. They're connected through unknown,
> >> extinct ancestors.
> >
> > Yes, like horses and tapirs. It would be interesting to know whether
> > the tree of Tapiroidea is as good as that of Equioidea.
> >
> > And Plato would agree with you that the extinct ones making the
> > direct connection are "unknown" but would add that Kathleen
> > Hunt probably had the "right opinion" about the ones connecting
> > a species of Hyracotherium to Equus.
>
> Is there any reason to recruit Plato into this discussion? If so, you
> must explain the relevance.

Socrates draws a sage distinction between knowledge, which is
impossible in biology, and right opinion, which we all believe
to occur from time to time. You might do well to read his
way of drawing the distinction.

> > Reference: Plato's Socratic dialogue "Meno."
> >
> > Hunt's sequence has:
> >
> > Hyracotherium/Eohippus - Orohippus - Epihippus - Mesohippus -
> > Miohippus - Parahippus - Merychippus - Dinohippus - Plesippus - Equus.
> >
> > One possible improvement is to put Kalobatippus between
> > Parahippus and Merychippus, but its fossils were too fragmentary
> > at the time of writing to be reasonably confident.
> >
> >> That is, there are nodes for which we have no data,
> >> and any method attempting to find relationships must take that into
> >> account. Every ancestral species is such a node.
> >
> > That last sentence is ideology, not biology.
>
> I think you need to stop with the pejorative terms if you want to keep
> this discussion civil.

It is you who are being uncivil here. Do you really think it
is biologically sound to say that we have no data for any ancestral
species?

> I'm willing to modify that to "The overwhelming
> number of ancestral species are such nodes and we have no reliable way
> to determine which ones are not".

That is completely different from the dogmatic claim you made.
Don't you realize that?

Peter Nyikos

[John Harshman]

On 4/15/16 6:13 PM, Peter Nyikos wrote:
> On Friday, April 15, 2016 at 8:00:32 PM UTC-4, John Harshman wrote:
>> On 4/15/16 4:54 PM, Peter Nyikos wrote:
>>> On Friday, April 15, 2016 at 7:27:14 PM UTC-4, John Harshman wrote:
>>>> On 4/15/16 4:08 PM, ruben safir wrote:
>>>>> On 04/15/2016 10:05 AM, John Harshman wrote:
>>>>>> are you proposing to connect species to species without any intervening,
>>>>>> internal nodes?
>>>>>
>>>>> there might be other species in between, but yes. Species are nodes and
>>>>> between species are edges.
>>>>>
>>>> Might I suggest that this is biologically and paleontologically
>>>> incorrect?
>>>
>>> You may suggest it, but you would only be right if you
>>> confined yourself to saying that this is incorrect
>>> according to the reigning cladistic ideoology.
>>
>> Sorry, but I don't think you have any idea what ruben/Popping is
>> proposing here
>
> Tell me how the context alters the plainly evident meaning of
> the words he wrote, which seems to talk about what Kathleen
> Hunt did, only with genera instead of species.

If you think there's a plainly evident meaning to what he said, I think
that's evidence you haven't read it carefully.

>>>> We don't know about the great majority of species that have
>>>> ever existed and have no fossils of them.
>>>
>>> So? If we have a fairly complete sequence of fossils, like
>>> the species-to-species linkages in some lines of the tree
>>> of Equidae, then one is perfectly justified in doing what
>>> Ruben said. Kathleen Hunt could have gone a good bit further than
>>> she did in filling out the tree in her article in the
>>> Talk.Origins Archive, but she restricted herself to
>>> depicting genera at the nodes of her tree.
>>>
>>> http://www.talkorigins.org/faqs/horses/horse_evol.html
>>>
>>> If you read her article, you can see that she could
>>> have listed a number of species at various nodes.
>>
>> She could have, but would she be right?
>
> Are you sure she would be wrong in even ONE instance?

No. I just think there's no way to tell if she was right in even one
instance.

>> Should you assume that you have
>> sampled all the relevant nodes? How is that good practice?
>>
>>>> This is even more obvious if
>>>> we're trying to find the relationships of extant species. They are not
>>>> connected through chains of fossils. They're connected through unknown,
>>>> extinct ancestors.
>>>
>>> Yes, like horses and tapirs. It would be interesting to know whether
>>> the tree of Tapiroidea is as good as that of Equioidea.
>>>
>>> And Plato would agree with you that the extinct ones making the
>>> direct connection are "unknown" but would add that Kathleen
>>> Hunt probably had the "right opinion" about the ones connecting
>>> a species of Hyracotherium to Equus.
>>
>> Is there any reason to recruit Plato into this discussion? If so, you
>> must explain the relevance.
>
> Socrates draws a sage distinction between knowledge, which is
> impossible in biology, and right opinion, which we all believe
> to occur from time to time. You might do well to read his
> way of drawing the distinction.

I really don't want to. Perhaps you can summarize and apply it to the
current case.

>>> Reference: Plato's Socratic dialogue "Meno."
>>>
>>> Hunt's sequence has:
>>>
>>> Hyracotherium/Eohippus - Orohippus - Epihippus - Mesohippus -
>>> Miohippus - Parahippus - Merychippus - Dinohippus - Plesippus - Equus.
>>>
>>> One possible improvement is to put Kalobatippus between
>>> Parahippus and Merychippus, but its fossils were too fragmentary
>>> at the time of writing to be reasonably confident.
>>>
>>>> That is, there are nodes for which we have no data,
>>>> and any method attempting to find relationships must take that into
>>>> account. Every ancestral species is such a node.
>>>
>>> That last sentence is ideology, not biology.
>>
>> I think you need to stop with the pejorative terms if you want to keep
>> this discussion civil.
>
> It is you who are being uncivil here. Do you really think it
> is biologically sound to say that we have no data for any ancestral
> species?

Yes, almost. We may have data for some ancestral species. We just have
no way to recognize which, if any, they are. Not sure why that would be
uncivil.

>> I'm willing to modify that to "The overwhelming
>> number of ancestral species are such nodes and we have no reliable way
>> to determine which ones are not".
>
> That is completely different from the dogmatic claim you made.
> Don't you realize that?

No. It's slightly different. And "dogmatic" is once more uncivil. Try
and cut down on adjectives. So, would you agree with my modified statement?

[ruben safir]

On 04/15/2016 07:27 PM, John Harshman wrote:
> Might I suggest that this is biologically and paleontologically
> incorrect? We don't know about the great majority of species that have
> ever existed and have no fossils of them.

I'll repeat that the spae of the relationships is not of interest when
doing a shortest path analysis. You can have one species, many or an
infinite number of species between two species. You keep thinking what
graphics algorithms build a data structure, and they do NOT, no more
than addition or multiplication does. They analyze the relationship
between objects in ***any existing structure***. That is the core
observation of the mathematical theories involving graphical
representations, which is that the specific representation is often
irrelevant to describe the relationship between objects, which in this
case is species.

So if, let say, you have a large number of primates. You should be able
to describe the individuals characteristics between two species, and
then a group of species, evaluating A/P or even generics, quantify these
characteristics as a 'cost' and then using theory like the shortest
path, determine with a great deal of certainty the relationship between
them, even without knowing the ancestral species, discovering how they
relate to each other.

I can see this discussion is creating some heat, and that is not my
intent. These algorithms, and other aspects of discrete mathematics,
are used in everything today for all kinds of analysis from facial
recognition to language recognition.

I thought they could be helpful to Paleontology, but I suppose I've been
shown to be wrong.


Ruben

[ruben safir]

On 04/15/2016 10:09 AM, John Harshman wrote:
> It seems here that all you're doing is calculating pairwise distances.
> And you are redefining "relationships" to be something entirely

> non-biological. Don't expect to interest any biologists in that, or even
> any mathematicians interested in biology.


ultimately you calculate all the pairings and order them by degrees of
closeness.

So if you have three species

a->b is 3
b->c is 2
a->d is 8

and e is related as follows

b->e 9
c->e 3
d->e is 2

using different algorithms you can find the shortest path from a->e
which might go through b, or b and c, or just d...

or maybe something else if the data is is not conclusive.

[John Harshman]

On 4/15/16 6:53 PM, ruben safir wrote:
> On 04/15/2016 07:27 PM, John Harshman wrote:
>> Might I suggest that this is biologically and paleontologically
>> incorrect? We don't know about the great majority of species that have
>> ever existed and have no fossils of them.
>
>
> I'll repeat that the spae of the relationships is not of interest when
> doing a shortest path analysis.

"spae"??

> You can have one species, many or an
> infinite number of species between two species. You keep thinking what
> graphics algorithms build a data structure, and they do NOT, no more
> than addition or multiplication does. They analyze the relationship
> between objects in ***any existing structure***. That is the core
> observation of the mathematical theories involving graphical
> representations, which is that the specific representation is often
> irrelevant to describe the relationship between objects, which in this
> case is species.
>
> So if, let say, you have a large number of primates. You should be able
> to describe the individuals characteristics between two species, and
> then a group of species, evaluating A/P or even generics, quantify these
> characteristics as a 'cost' and then using theory like the shortest
> path, determine with a great deal of certainty the relationship between
> them, even without knowing the ancestral species, discovering how they
> relate to each other.

I don't think you're using a lot of words the same way they're used by
other people, and you're using them very loosely and ambiguously too.
Consequently, I don't know what you're proposing to do.

> I can see this discussion is creating some heat, and that is not my
> intent. These algorithms, and other aspects of discrete mathematics,
> are used in everything today for all kinds of analysis from facial
> recognition to language recognition.
>
> I thought they could be helpful to Paleontology, but I suppose I've been
> shown to be wrong.

Hard to tell. What do you think those algorithms would contribute? What
would we learn? What is the biological interpretation of your "shortest
path"?

[John Harshman]

Three problems:

1. What does "shortest path" mean biologically"

2. What does "degrees of closeness" mean biologically?

3. Why should a path from one species to another go through a third
species, and what would that mean biologically?

[RSNorman]

On Fri, 15 Apr 2016 21:53:07 -0400, ruben safir <ru...@mrbrklyn.com>
wrote:

I think the easiest way to get at the issue is perhaps to ask the
question:

Yes, your algorithms can generate the "shortest path". What
significance does that shortest path have to biology?

[RSNorman]

Sorry about my post on exactly this. I was on the road all day
yesterday and am only now slowly going through the posts.

The lesson is to read them all before responding to any.

[RSNorman]

On Fri, 15 Apr 2016 11:41:18 -0700, John Harshman
<jhar...@pacbell.net> wrote:

>On 4/15/16 11:25 AM, Popping mad wrote:
>>
>> that just means they didn't do it correctly. The suggestion made
>> previously, for example about Lizzards, Birds and Crocs is laughable.
>
>I presume it was intended as an analogy only, not a serious suggestion.
>To my knowledge nobody has proposed a phylogeny in which birds and crocs
>are not both archosaurs. Nor do I recall any pheneticist doing an
>analysis of reptiles

It was not an analogy only.

It is a simple fact that for ages lizards and croc were both in one
class, Reptilia, while robins were in a different class, Aves. The
tree of classical taxonomic ranks is a nested hierarchy. The problem
was that the classical taxonomy was not a phylogeny. It is true that
one time lizards and crocs were "grouped more closely" than birds and
crocs.

[John Harshman]

However, this classification persisted long after we became interested
in phylogenies. Further, it isn't completely clear that if you
classified based on overall similarity crocs would not go with birds. It
all depends on which characters you look at (which is one reason
"overall similarity" isn't a popular measure). Now, if we only knew what
ruben was proposing.