Study on Improving Fluid Intelligence through Cognitive Training System Based on Gabor Stimulus

[zzzz]

http://www.computer.org/portal/web/csdl/doi/10.1109/ICISE.2009.1124

[Gwern Branwen]

On Fri, Mar 5, 2010 at 11:39 AM, zzzz <filip...@gmail.com> wrote:
> http://www.computer.org/portal/web/csdl/doi/10.1109/ICISE.2009.1124

My library doesn't have this. It seems a little suspicious to me: just
10 training sessions? And since when does the IEEE cover this sort of
thing?

But the real question is, does it improve on Jaeggi 2008 with full
time allotted for the Raven's, or is the time drastically curtailed?
If the latter, then the study would tell us nothing.

--
gwern

[cev]

What are Gabor stimuli?

[Jonathan Toomim]

A Gabor stimulus (or Gabor patch) is a sinusoidal grating masked by a
gaussian function. They're very simple visual stimuli, and are the
stimulus type that the so-called "simple cells" in V1 (primary visual
cortex) are optimally tuned for. They're very commonly used in vision
science and with experimental computer vision models.

http://mrl.nyu.edu/~perlin/experiments/gabor/gabor2d.jpg

Jonathan

On Mar 5, 2010, at 2:25 PM, cev wrote:

> What are Gabor stimuli?
>
> --
> You received this message because you are subscribed to the Google
> Groups "Dual N-Back, Brain Training & Intelligence" group.
> To post to this group, send email to brain-t...@googlegroups.com.
> To unsubscribe from this group, send email to brain-trainin...@googlegroups.com
> .
> For more options, visit this group at http://groups.google.com/group/brain-training?hl=en
> .
>

[Jonathan Toomim]

My library doesn't either. Bummer.

They don't mention a control group in the abstract. If there's no
control group, this study is worthless—all it would mean then is that
people who take the Raven's twice do slightly better the second time.

These are just conference proceedings. Until they publish their
findings in a peer-reviewed journal, I'm going to ignore this.

[KD Jones]

Did Valdeane Brown have anything to do with this? He's got a Gabor
thing going on.

- KD

[cev]

Thanks, Jonathan

[KD Jones]

Is this a source?

http://www.wipo.int/pctdb/ja/wo.jsp?WO=2007070876&IA=US2006062175&DISPLAY=DESC

[KD Jones]

(No, I don't think that was a source. Appeared in the end to have
something to do with Posit Science stuff.)

I just spent two hours trying every back door I could find, attempting
to figure out what sort of task they were using.

Nothing.

But, it was somewhat educational for my little hick brain. With
regard to related studies the number of references to terms related to
"visuospatial" work (hence gabor stimuli) was astounding - even when I
stopped drilling with "gabor" and searched more directly toward
"intelligence."

It's incredible to me how related fields are touching each other now,
even having started out aware of the convergences. I'm very thankful
I have people who actually know something (y'all) to point this stuff
out.

[david sky]

There appears to be some confusion here. The Gabor Stimulus referenced
in this study is Dual-N-Back Training as conducted by Jaeggi et al
(2008). Sorry to throw a monkey wrench in the discussion.

[Gwern Branwen]

My current opinion on Qiu et al 2009 is that it was poorly done and
perhaps even fraudulent.

I pointed out earlier that Chinese science does not have a very good
reputation and issues are rampant. Reading Qiu for the meta-analysis,
I noticed that the graph of the final IQ scores exhibited a bizarre
behavior where the control group's IQ apparently dropped 10 points. I
say apparently because the actual numbers weren't included; I emailed
the authors to ask about this and for the means, _n_s, and SDs, but my
5 emails to the 4 authors since May have all gone unanswered (no
bounces or other error messages).

--
gwern
http://www.gwern.net

[Michael]

no reason to throw coins into their little hat then...

[☉]

We should probably devote our attention to research institutes and
researchers with reputable backgrounds. However, systematic fraud
(instead of isolated cases), across country borders, will pose a
problem for this selective weighting. Even psychology as a whole
suffers from this situation, such that one can put up the case of
parapsychology and draw numerous analogies therebetween, suggestive of
the credulousness of those who believe in *untenable demarcation
criteria* to prefer one field over another qua "science" instead of
"pseudoscience", sensu showing with "statistical
significance" (perhaps equivocally, it may be argued) that that
field's respective phenomena have been shown to have their specific
effect (minus any background theory to predict this outcome). As such,
I believe – however tedious and resistant to heuristic analysis – that
we should proceed on a case-by-case basis with any criticism
pertaining to this area of inquiry (if it is to be genuinely deemed as
one worthy of the predicate "scientific", or as one upholding a
worthwhile understanding of a scientifically motivated methodology).

We should not want to enlarge unfair attention on this particular
case. Indeed, gwern, you haven't conveyed with much rigor the actual
reasons (justifiably) that there was either poor conduct or fraud
here. Their unresponsiveness cannot be construed as evidence (that
would be to argue from silence); and the "bizarre behavior" you
describe may not exclusively describe either situation, but rather be
symptomatic of the null hypothesis itself (or even something under the
catch-all hypothesis). What logical grounds do you actually have for
saying this? If you have none, it would perhaps be wise to remain
uncommitted to the legitimacy of the study, and nevertheless state
specific concerns, rather than arguing for the positive case of fraud
or poor conduct, which hasn't in the least been proved here.

argumzio

On Jul 27, 10:15 am, Gwern Branwen <gwe...@gmail.com> wrote:
> On Fri, Mar 5, 2010 at 11:56 AM, Gwern Branwen <gwe...@gmail.com> wrote:

[Gwern Branwen]

On Fri, Jul 27, 2012 at 1:21 PM, ☉ <argu...@gmail.com> wrote:
> Their unresponsiveness cannot be construed as evidence (that
> would be to argue from silence)

Of course arguments from silence are valid. If fraudulent authors are
less likely to reply than non-fraudulent authors, then no reply must
be some degree of evidence for fraud; similarly for poorly done.

Qiu's 4 authors are the only ones to not reply at all (with the
exception of Norbert Jaušovec, who I contacted for literature
suggestions & have not followed up on yet because his paper provided
all the necessary data in the first place).

> and the "bizarre behavior" you describe may not exclusively describe either situation, but rather be symptomatic of the null hypothesis itself (or even something under the catch-all hypothesis).

A drop of 10 or more points, overcoming testing effects and other
things, in any sort of sample size, is very suspicious even under the
null hypothesis.

--
gwern
http://www.gwern.net

[☉]

On Jul 27, 12:38 pm, Gwern Branwen <gwe...@gmail.com> wrote:
> Of course arguments from silence are valid.

It may be argued that they are in some cases, but validity here should
mean "in all cases"; and you're far from even demonstrating that any
given argument from silence (a well-known fallacy) is valid in all
cases based on this single instance, let alone "of course" (as if it
were valid a priori).

> If fraudulent authors are
> less likely to reply than non-fraudulent authors, then no reply must
> be some degree of evidence for fraud; similarly for poorly done.

Yes, but that doesn't mean that: if there is no reply, then the
authors are more likely to be fraudulent than non-fraudulent. Please
do not affirm the consequent here.

> Qiu's 4 authors are the only ones to not reply at all (with the
> exception of Norbert Jaušovec, who I contacted for literature
> suggestions & have not followed up on yet because his paper provided
> all the necessary data in the first place).

An exception does not provide evidence of fraud in this case.

> A drop of 10 or more points, overcoming testing effects and other
> things, in any sort of sample size, is very suspicious even under the
> null hypothesis.

"Suspicious" is very different from saying "poorly done" or
"fraudulent".

Please, I'm asking for serious logical grounds for your claim, not
circumstantial (inadequate) inductive evidence which can be molded
however one pleases. Note: I am not saying that this isn't a case of
fraud; I'm asking you if you have positive proof (evidence) that this
is a case of fraud, deliberate misconduct, or incompetence (the third
which you seem to neglect as a viable possibility – indicative of a
false dichotomy).

argumzio

[☉]

On Friday, July 27, 2012 2:29:14 PM UTC-5, ☉ wrote:

On Jul 27, 12:38 pm, Gwern Branwen <gwe...@gmail.com> wrote:
> If fraudulent authors are
> less likely to reply than non-fraudulent authors, then no reply must
> be some degree of evidence for fraud; similarly for poorly done.

Yes, but that doesn't mean that: if there is no reply, then the
authors are more likely to be fraudulent than non-fraudulent. Please
do not affirm the consequent here.

Your statement can further be boiled down to this: "if fraud is correlated with (evidence of) silence (or if non-fraud is inversely correlated with silence), then silence must be evidence of (correlated with) fraud" which is a case of petitio principii so laughable and ridiculous that I wonder why I didn't notice immediately. Perhaps because I recently had lunch and it's been a slow afternoon. In short, fraud is not equivalent to silence.

argumzio

[Gwern Branwen]

On Fri, Jul 27, 2012 at 3:29 PM, ☉ <argu...@gmail.com> wrote:
> Please, I'm asking for serious logical grounds for your claim, not
> circumstantial (inadequate) inductive evidence which can be molded
> however one pleases. Note: I am not saying that this isn't a case of
> fraud; I'm asking you if you have positive proof (evidence) that this
> is a case of fraud, deliberate misconduct, or incompetence (the third
> which you seem to neglect as a viable possibility – indicative of a
> false dichotomy).

I do not have deductive syllogistic grounds for fraud or sloppiness,
and likely never will; for example, Simonsohn who has successfully
uncovered cases of scientific fraud which I'm sure we have all heard
of, mentions in his discussion of his methods
(http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2114571 "Just Post
It: The Lesson from Two Cases of Fabricated Data Detected by
Statistics Alone") that he strongly believes he found several other
cases but the authors say they lost the data or ignore him or
whatever; how will he ever prove his case? For one paper, the data was
lost not long after requests were made... Trapped in a deductive
framework, only outright confession will ever prove him right. Such
logic is not useful in the real world, and I don't understand your
insistence on pushing scientific problems into a highly limited narrow
framework they have never fit into well. It would be helpful if in the
future you labeled your criticisms 'useless pointing out that
something would be a deductive fallacy' and 'possibly useful and
relevant inductive or statistical objection'.

On Fri, Jul 27, 2012 at 3:43 PM, ☉ <argu...@gmail.com> wrote:
> Your statement can further be boiled down to this: "if fraud is correlated
> with (evidence of) silence (or if non-fraud is inversely correlated with
> silence), then silence must be evidence of (correlated with) fraud" which is
> a case of petitio principii so laughable and ridiculous that I wonder why I
> didn't notice immediately.

Well, yes, that's the whole point: if you grant a correlation in one
direction, you must grant a correlation in the other. -_- If we grant
honest researchers are more likely than dishonest researchers to reply
to an inquiry about problems in their papers & for their data, then we
must also grant that dishonest researchers are more likely to not
reply. If a reply => ~fraud, then the contrapositive is fraud =>
~reply. And from there, we make the additional assumption that fraud
is the only sufficient reason for ~reply, and then we can affirm the
consequent and conclude from ~reply, fraud. The generalization to a
statistical interpretation where there are multiple sufficient reasons
for ~reply (some reasons innocent, like no longer using the email
addresses) is pretty straightforward.

--
gwern
http://www.gwern.net

[☉]

On Jul 27, 2:55 pm, Gwern Branwen <gwe...@gmail.com> wrote:
> I do not have deductive syllogistic grounds for fraud or sloppiness,
> and likely never will;

Thanks. As long as we both understand that, and you don't use "valid"
w.r.t. any argument from silence.

> for example, Simonsohn who has successfully
> uncovered cases of scientific fraud which I'm sure we have all heard
> of, mentions in his discussion of his methods
> (http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2114571"Just Post
> It: The Lesson from Two Cases of Fabricated Data Detected by
> Statistics Alone")

Sleuthing is an incredibly important process, and one with which I am
personally familiar. Yes, I have heard of him.

> that he strongly believes he found several other
> cases but the authors say

>they lost the data or

If we're to give them the benefit of the doubt, then that doesn't
imply fraudulent activity, but rather incompetence.

>ignore him or

Who in this world hasn't been ignored at least once?

> whatever;

It would be a good idea to make a list of all the kinds of resonses
and non-responses he's received. "Whatever" isn't satisfactory.

> how will he ever prove his case?

By claiming "this is fraud" and receiving no reply from whom he
accuses, presumably.

> For one paper, the data was
> lost not long after requests were made...

Incompetence.

> Trapped in a deductive
> framework, only outright confession will ever prove him right.

If he limits himself to finding "fraud" on such a view, yes.

> Such
> logic is not useful in the real world,

You're quite wrong, because it allows us to be clear about what we
say. I prefer clarity and rigor to sloppiness in expression. Besides,
it's unfair if there's no colloquium to which they're said to belong
to police these things. What are the explicit criteria for "fraud"?
What minimum requirements must be met so that one doesn't get branded
a crank? These things must be addressed.

> and I don't understand your
> insistence on pushing scientific problems into a highly limited narrow
> framework they have never fit into well.

It's not a framework. It's clarity and understanding. I haven't seen
much on this point.

> It would be helpful if in the
> future you labeled your criticisms 'useless pointing out that
> something would be a deductive fallacy' and 'possibly useful and
> relevant inductive or statistical objection'.

In the future, it would be helpful for you not to make claims not
justified by the evidence. And negative evidence is the least
worthwhile candidate to meet a positive measure of something.

argumzio

[☉]

On Friday, July 27, 2012 3:06:00 PM UTC-5, ☉ wrote:

In the future, it would be helpful for you not to make claims not
justified by the evidence. And negative evidence is the least
worthwhile candidate to meet a positive measure of something.

I forgot mention in closing that: if you cannot acquire the raw data from them, then it is apparent you'll be unable to conduct any worthwhile analysis. Hence, you're better off not including the paper as given - i.e., it's worthless junk.

argumzio

[jotaro]

i love how argumzio is so critical its almost out of this world

like a lawyer.

do you really feel so much need to be so critical about everything because it can be quite time consuming.

i hope you are critical to things that actually really matter and not towards everything thats gets in the way .

care to tell me more how you got in such a position? i am preety curious about it.

 

[☉]

On Friday, July 27, 2012 5:06:42 PM UTC-5, JokyBoy wrote:

i love how argumzio is so critical its almost out of this world

like a lawyer.

Lawyers aren't critical. They're paid sophists. ;)

 

do you really feel so much need to be so critical about everything because it can be quite time consuming.

I suppose there is little need, more like a bad habit I can't quit. Yes, it is time-consuming, but on occasion I have free time to let my regnant habits have their say.
 

i hope you are critical to things that actually really matter and not towards everything thats gets in the way .

"...that actually really matter..." - yes, I would have hoped that, this being the brain-training, intelligence, DNB group that it actually really mattered to call someone out on an untenable claim of fraud by a third party.
 

care to tell me more how you got in such a position?

Why is the sky blue and why does crap stink?

 argumzio

[☉]

I thought I made it clear to your attention-starved existence that I won't respond to you privately. Go play a video game or something.

argumzio

On Fri, Jul 27, 2012 at 9:00 PM, shuriken <m.vel...@gmail.com> wrote:

can you clarify this for me you: have add, don't you? it's a neutral observation, no pun intended.

On Friday, July 27, 2012 9:29:14 PM UTC+2, ☉ wrote:

On Jul 27, 12:38 pm, Gwern Branwen <gwe...@gmail.com> wrote:
> Of course arguments from silence are valid.

It may be argued that they are in some cases, but validity here should
mean "in all cases"; and you're far from even demonstrating that any
given argument from silence (a well-known fallacy) is valid in all
cases based on this single instance, let alone "of course" (as if it
were valid a priori).


> If fraudulent authors are
> less likely to reply than non-fraudulent authors, then no reply must
> be some degree of evidence for fraud; similarly for poorly done.

Yes, but that doesn't mean that: if there is no reply, then the
authors are more likely to be fraudulent than non-fraudulent. Please
do not affirm the consequent here.


> Qiu's 4 authors are the only ones to not reply at all (with the
> exception of Norbert Jaušovec, who I contacted for literature
> suggestions & have not followed up on yet because his paper provided
> all the necessary data in the first place).

An exception does not provide evidence of fraud in this case.


> A drop of 10 or more points, overcoming testing effects and other
> things, in any sort of sample size, is very suspicious even under the
> null hypothesis.

"Suspicious" is very different from saying "poorly done" or
"fraudulent".

Please, I'm asking for serious logical grounds for your claim, not
circumstantial (inadequate) inductive evidence which can be molded
however one pleases. Note: I am not saying that this isn't a case of
fraud; I'm asking you if you have positive proof (evidence) that this
is a case of fraud, deliberate misconduct, or incompetence (the third
which you seem to neglect as a viable possibility – indicative of a
false dichotomy).

argumzio

[☉]

I didn't reply to the following portion of gwern's earlier message because of time-constraints, so here it is now.

On Friday, July 27, 2012 2:55:32 PM UTC-5, gwern wrote:

Well, yes, that's the whole point: if you grant a correlation in one
direction, you must grant a correlation in the other. -_-

Assuming that one actually granted it, yes. :)

 

If we grant
honest researchers are more likely than dishonest researchers to reply
to an inquiry about problems in their papers & for their data, then we
must also grant that dishonest researchers are more likely to not
reply.

I don't see how anyone can seriously entertain this as a real empirical phenomenon, unless it is merely an informal heuristic. And heuristics aren't always assured of being right... but let's leave that aside.

 

If a reply => ~fraud, then the contrapositive is fraud =>
~reply.

Yes, but you're trying to say that ~R -> F, which isn't even valid based on what you've just granted, and ~R is all you have to go on, so proof by contraposition fails to prove F. Further, if ~R -> F were true, then you'd have to grant ~F -> R. Are you sure you would want to say these things in addition to the above?
 

And from there, we make the additional assumption that fraud
is the only sufficient reason for ~reply

Indeed, I pointed out you'd assumed what you're trying to prove, that is, that ~R always follows from F (but this is merely to mischaracterize the situation as saying that ~R occurs strictly only when F is true, i.e., F&~R, not F->~R), which is shown by the way in which you'd confused what you said with what you wanted to say. Moreover, ~R has, according to the paradox of material implication, an infinite variety of sufficient conditions which imply it, so it cannot be a strict implication based on F->~R alone. Straightforward commonsense could also show that silence can follow for any variety of reasons, e.g., why I didn't respond to this part of your post until now. You said:

F -> ~R and so ~R -> F

Or paraphrasing you again: if fraud is evidence of (implies) silence [if F->~R], then silence implies fraud [then ~R->F]. Do you see your error in this highly condensed paraphrase? -_-

You didn't execute the contrapositive correctly in the original English wording, but you did it correctly once I alerted you to it; however, you don't seem yet to realize why I said what I did in the first place. :)
 

, and then we can affirm the
consequent and conclude from ~reply, fraud.

This is exactly what you did wrong. From "F -> ~R" one can only conclude "R -> ~F" not "~R -> F". This is why I said one cannot treat F <=> R (i.e., that F->~R & ~R->F) as true based on the initial assumption of F->~R.
 

The generalization to a
statistical interpretation where there are multiple sufficient reasons
for ~reply (some reasons innocent, like no longer using the email
addresses) is pretty straightforward.

Let me break it down into the idealized case for you. Here are your two assumptions (givens):

1. ~R
2. F->~R

How do you conclude F from these? Answer: we cannot.

From two I can get the following by contraposition:

3. R->~F

So, you make the error (assumption) that ~R->F, let's call that 4.

Then you see, you don't even need F->~R, because from 1 and 4 one can conclude F by a simple application of modus ponendo ponens. But you clearly can't get 4 from 2 or 3. Do you have any justification for making the additional assumption 4 and ignoring lines 2 and 3? You used the example of Simpsonsohn as a justification for 2 (and so 3), which anyone can grant inductively. However, entities should not be multiplied unnecessarily, (while your inductive evidence you apparently yourself ignored in the proof).

For the real-world case, in which the premisses always have a value less than one (meaning that they aren't always true), we do not have any fundamental difference in the analysis of this: it is still erroneous. That is, from ~R, we can obtain any implication we please, besides F.

Logic is an aid to clarity, and helps us from making errors which could be avoided when it is followed rigorously. It should not be spurned, even when the premisses are probabilities less than unity.

argumzio
 

[Aman Abdullahi Idle]

Lol harsh. 

Sent from my iPhone

--

You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.

To view this discussion on the web visit https://groups.google.com/d/msg/brain-training/-/93JNFI39u6wJ.

[Michael]

Do you read non-fiction books, and, enjoy them?

You don't have to answer if you don't want to; I won't make you a liar out of it, promise.

;)

[☉]

Testimonials are worthless. That said, on some days I read non-fiction books and enjoy them, on some days I do not enjoy them, and on all other days I do not read non-fiction books.

I can say yes and no and wouldn't be a liar at all. ;)

argumzio

[Michael]

Sorry, I meant to say *fiction* books. Rather silly in fact, for me to ask if you have am inclination towards non-fiction books, any old-fellow here could above-all see that this would be a tendency that perhaps has a greater chance of being favored compared to others. One may even say that it was so silly that I may have been mocking you, but no, definitely not.

It was a very long day for me; the value of testimonials aside, I had not slept much at all and wine had been too much of a good friend the night before. On top of that, given my history here, one might say that I'm prone to making errors relating to small details; more than anything else anyway.

[Gwern Branwen]

On Fri, Jul 27, 2012 at 10:52 PM, ☉ <argu...@gmail.com> wrote:
> I don't see how anyone can seriously entertain this as a real empirical
> phenomenon, unless it is merely an informal heuristic. And heuristics aren't
> always assured of being right... but let's leave that aside.

'merely'? A heuristic *is* an empirical phenomenon.

> Moreover, ~R has, according to the paradox of
> material implication, an infinite variety of sufficient conditions which
> imply it, so it cannot be a strict implication based on F->~R alone.
> Straightforward commonsense could also show that silence can follow for any
> variety of reasons, e.g., why I didn't respond to this part of your post
> until now.

Missing the point about sole sufficient and the generalization.

--
gwern
http://www.gwern.net

[☉]

On Jul 28, 3:09 pm, Gwern Branwen <gwe...@gmail.com> wrote:

> On Fri, Jul 27, 2012 at 10:52 PM, ☉ <argum...@gmail.com> wrote:
> > I don't see how anyone can seriously entertain this as a real empirical
> > phenomenon, unless it is merely an informal heuristic.
>

> 'merely'? A heuristic *is* an empirical phenomenon.

Merely does not change the logical status of the statement. Perhaps
you need to look into the function of "unless".

> > Moreover, ~R has, according to the paradox of
> > material implication, an infinite variety of sufficient conditions which
> > imply it, so it cannot be a strict implication based on F->~R alone.
> > Straightforward commonsense could also show that silence can follow for any
> > variety of reasons, e.g., why I didn't respond to this part of your post
> > until now.
>
> Missing the point about sole sufficient and the generalization.

You're either missing the point of the point (where PMI was only
ancillary to the point) or being facetious.

Either way, I'm quite finished and have put the whole thing behind
me. :)

argumzio

[☉]

On Jul 28, 12:39 pm, Michael <dicone...@gmail.com> wrote:
> Sorry, I meant to say *fiction* books.

It may or may not be self-obvious, but the same response could be
given to "fiction books" as well.

However, I question the relevance of the question, so I do not
particularly care to inject this kind of digression into the thread.

argumzio

[EleazarWheelock]

Neither syllogistic or first order sentential logic is particularly clarifying in this situation. 

Perhaps if we were attempting to deductively demonstrate "positive proof" it would be. But we're not. We're talking about evidence. 

Seems to me Gwern is arguing from a Bayesian perspective. As such, he'd be correct in claiming that arguments from silence are valid. 

The reasoning can be made explicit with conditional probability: 

P0. Let's define evidence (X) as a prior which increases the probability of some epistemic claim (Y). Thus we are warranted in denoting X evidence

precisely in the case that P(Y|X) > P(Y|~X)  (where P(x|y) is read "The probability of x given y"). 

P1. P(~F|R) > P(~F|~R)                  

1. P(F|R) < P(F|~R)                           P1

2. P(F|~R) > P(F|R)

Let  r = ~R

3. P(F|r) > P(F|~r)       

Let F = fraud, R = reply, and r = nonreply

The probability of a fraud given a nonreply is greater than the probability of fraud given a reply. Ergo, a nonreply is evidence of fraud    P0

QED

This of course does not mean nonreplies are _strong_ evidence. Much less that nonreplies are stronger evidence of mischief than of, say, incompetence. But nonreplies do meet a minimum criterion of evidence. And in general the line of reasoning is itself inductively valid*.  

*The general case is proven more explicitly below: 

PI. P(B|A) > P(B|~A)       trans.    A is evidence of B,  where P(x|y) is read "The probability of x given y)

PII.  a = ~A                    trans.   a is the absence of evidence

PIII. b = ~B                    trans.   b is absence (of B)

1. P(B|A) = 1 - P(~B|A)     and       P(B|~A) = 1 - P(~A|~B)          (Complements)

2.  1 - P(~B|A) > 1 - P(~B|~A)                                                    (Substitution)

3.  P(~B|A) < P(~B|~A)                                                              (AM)

4.  P(b|~a) < P(b|a)                                                                    (Substitution, ~a = ~~A = A)

From I, we conclude "a (absence of evidence) is evidence of b (absence)"

QED

[Michael]

It may or may not be self-obvious but the same response could also be given about food, sport, clothing, movies and many other things to boot, but this situation is probably unlikely to arise, on average, given the assumption that 1) people have the liberty to place different levels of consideration on each category, and, 2) People generally don't make similar approximations about different things, especially as the list grows in number, 

These are the assumptions I probably shouldn't have made, my apologies. Maybe all people really are green!

Considering no body likes potholes in the road, feel free to drive around this post instead of partaking in a changing of books (otherwise known as: digression), as I suggested previously,  thus avoiding any unnecessary reply, like before.

[hallu]

Let's be unreasonably generous, 99 out of 100 frauds will be silent, P(S|F)=0.99. Does such a high probability concurrently imply a comparable probability of silent people being frauds, P(F|S)?

Given P(S|F)=0.99 and P(F)=0.000056*, P(F|S) will fall around 0.56%*.

*Retractions in the scientific literature: is the incidence of research fraud increasing?, Steen 2010, http://www.sendspace.com/file/6jzyu0

*http://en.wikipedia.org/wiki/Base_rate_fallacy

[Gwern Branwen]

Indeed; given how little information we have, and *can* have, about
any particular paper, I'd regard 1% as a useful point to mention in
the absence of better information. But I'd also note that the base
rate for fraud is a lot higher for Chinese research. Whether this more
than counterbalances the overly generous assumption about how
diagnostic silence is and gives a higher posterior than 1%, I leave to
readers.

--
gwern
http://www.gwern.net

[☉]

You completely miss the mark in that you fail to realize your question was a very bad question and would have never in a million years led me to give you the information you apparently hoped for me to have provided (viz., any question that can be answered with both "yes" and "no" is not a good one).

argumzio

[☉]

On Monday, July 30, 2012 1:49:18 AM UTC-5, EleazarWheelock wrote:

Neither syllogistic or first order sentential logic is particularly clarifying in this situation.

Yes, it is, for any syllogistic procedure would reveal the fallaciousness by which the conclusion was reached. To deny this is tantamount to obscurantism. :)
 

Perhaps if we were attempting to deductively demonstrate "positive proof" it would be. But we're not. We're talking about evidence. 

"Positive proof" is all we should be concerned with. The evidence provided is not any kind of evidence at all. In other words, the evidence is incoherent. If you seriously believe non-evidence is evidence, then we might as well consider all of the non-evidence and spuriously conceive ways of making it seem evidential for anything whatever. ;)
 

Seems to me Gwern is arguing from a Bayesian perspective.

Were that the case, he would have 1) said so, 2) provided his prior probabilities, and 3) actually shown the process by which he raised his priors on account of the non-reply (though such non-reply criteria potentially expose something unsound in the whole procedure). Maybe by my not responding to anything I should be saying all sorts of things, on your view, but I would never grant that in this lifetime or the next. By the way, parapsychologists make use of Bayesian "reasoning" all of the time – a fact which may change your priors about certain things qua priors and posteriors. :)
 

As such, he'd be correct in claiming that arguments from silence are valid. 

"Correct" only w.r.t. to Bayesian reasoning, which is a presumption I'm not willing to grant – with good reason, no doubt! No, "valid" means one thing here, and it certainly doesn't apply to anything pertaining to Bayesianism.

As such, your explication is utterly worthless, but I may go over it in detail when I have the time. This is positive evidence that I will consider it, given sufficient time. :) 

From I, we conclude "a (absence of evidence) is evidence of b (absence)"

QED

"Quite easily done", indeed. By the way, what you just proved (that is, that which you've put forward prior to any given refutation), as noted within the quotation marks, has nothing to do with the notion that absented evidence is positive evidence of something else (besides the absence w.r.t. the evidence). Perhaps you'd like to give us a positive proof of the Divinity via Bayesianism as well? :)

In any case, would that really live up to a serious standard of the scientific ethos? Not one in which I should find myself participating....

argumzio

[Gwern Branwen]

On Mon, Jul 30, 2012 at 3:57 PM, ☉ <argu...@gmail.com> wrote:
> Were that the case, he would have 1) said so, 2) provided his prior
> probabilities, and 3) actually shown the process by which he raised his
> priors on account of the non-reply (though such non-reply criteria
> potentially expose something unsound in the whole procedure).

Priors don't matter unless you want a specific posterior; the
'strength' of an observation can be expressed as a logarithmic factor
which you simply multiply against your prior (whatever that is).
Statisticians actually consider simply reporting priors & posteriors
after an experiment or observation to be bad form, since your readers
may well not share your exact prior but could use the calculated
logodds to update whatever their prior was. I didn't bring up priors &
posteriors because they were not relevant to the question of whether
the log factor was non-zero, and if non-zero, what the sign was.
Before we do priors and posteriors, we must do log factors, and before
we do that, we ought to know what sign the factor will have! If we
didn't get that far, it's because you refused to go that far, and it's
absurd of you to object now.

My point about Chinese base rates is a good example of why we might
want to split our calculation in this way; one can disagree with me
about the exact base rate we would use as a prior, but still agree
with the basic logarithmic calculation of how to update whatever base
rate we choose. So we pass from disagreeing completely about the
entirety of the end result, to agreeing about part of the calculation
but disagreeing about a specific part which we can productively
discuss further, and even if we continue to disagree, we can keep in
sync as we make future observations and agree on the log factors each
observation is worth.

Statistics! It's so much more useful than zeroth-order logic!

--
gwern
http://www.gwern.net

[EleazarWheelock]

1.  The simple fact is that outside the orbit of Vulcan most arguments which are rationally compelling are not deductively valid. 

This is painfully obvious. To say it's "tantamount to obscurantism" is patently untrue. Going back to Aristotle's Rhetorics, it's well known in the philosophical literature that deductive logic is of qualified utility when trying to account for inductive inference and scientific reasoning. Recognition of this is why we have an entire discipline called Defeasible Reasoning to assist with problems not amenable to deductive analysis. If non-deductive (aka deductively invalid) modes of reasoning like "inference to the best explanation, abduction, analogical reasoning, and scientific induction" have any merit--and they manifestly do--"positive proof" should hardly be our exclusive concern. 

2. Since when does Gwern have to explicitly claim he's arguing from a Bayesian perspective for me to legitimately say it is such? It's bloody obvious if you attend to the evidence. You might as well claim he must preface all cosmological claims with the gravitational constant and relevant Newtonian formulae in order to conclude he believes in gravity. What's more, he's welcome to take exception to my formalization of his reasoning. Ironically, I take the lack of evidence of his disagreement as evidence of the absence of disagreement! And even if he weren't consciously Bayesian, I can without loss of generality--and did--make an equivalent argument in Bayesian terms and thus it can rightly be termed a Bayesian perspective. Criteria 2 and 3 are also completely unnecessarily for this argument to be justifiably termed as coming from a general Bayesian perspective. 

3. Validity here "certainly doesn't apply to anything pertaining to Bayesianism"? Wrong again. Valid means exactly what it normally means. If you grant the truth premises 1, 2, & 3, the truth of the conclusion necessarily follows. If you wish to not assent to any premises---stated or implied (like the consistency or coherency of Bayesianism, arithmetic, probability theory, etc)--that's of course your prerogative. My point was simply to give a formal and consistent account of someone's reasoning. That's all. I'm quite serenely agnostic on the ultimate truth of the argument. But my prima facie opinion is that while silence is evidence, it's very weak evidence. Where the rubber meets the road, we are in broad agreement. 

4. "By the way, what you just proved (that is, that which you've put forward prior to any given refutation), as noted within the quotation marks, has nothing to do with the notion that absented evidence is positive evidence of something else (besides the absence w.r.t. the evidence)."  On one interpretation, this is exactly wrong. The proof is not tautologous. It does not say "absence of evidence is absence of evidence". But I'm not sure how you're using the term "positive evidence" or "something else". (E.g. "Absence of evidence for the Yeti is evidence for his (existential) absence." I'm afraid existential absence (nonexistence) is indeed quite "something else" than mere evidentiary absence.) Or what exactly you mean by "absence w.r.t to the evidence".  If by the latter you mean my conclusion is contingent on the premise of a defeasible claim, viz. the absence of evidence, then, yes. But that's perfectly obvious and non-fatal...Saying "absence of evidence is evidence of absence (conditioned on the absence of the evidence)" is just needlessly appending the premise to the conclusion with a pretense to insight. 

5. Perhaps you'd like to give us a positive proof of the Divinity via Bayesianism as well? :)" 

I'm not even sure it's coherent to talk of "positive proof" via Bayesianism. You have to be careful how you deploy "evidence" "positive evidence" and "proof". Bayesian claims are always subject to defeaters. It doesn't axiomatically define, declare and deduce unassailable "positive proofs". That's much of the point I've been trying to get across...

That said, if a proof of the Original Mind were forthcoming, I'd hope you'd have the humility to accede regardless of its humble provenance. 

"..good tidings may yet come from [utterly worthless] Nazareth." 

Until then, you'll have to be content with a more modest credo: There is no God. And argumzio is his prophet. 

[Gwern Branwen]

On Tue, Jul 31, 2012 at 6:09 PM, EleazarWheelock
<Allan.Carl....@gmail.com> wrote:
> 2. Since when does Gwern have to explicitly claim he's arguing from a
> Bayesian perspective for me to legitimately say it is such? It's bloody
> obvious if you attend to the evidence. You might as well claim he must
> preface all cosmological claims with the gravitational constant and relevant
> Newtonian formulae in order to conclude he believes in gravity. What's more,
> he's welcome to take exception to my formalization of his reasoning.
> Ironically, I take the lack of evidence of his disagreement as evidence of
> the absence of disagreement! And even if he weren't consciously Bayesian, I
> can without loss of generality--and did--make an equivalent argument in
> Bayesian terms and thus it can rightly be termed a Bayesian perspective.
> Criteria 2 and 3 are also completely unnecessarily for this argument to be
> justifiably termed as coming from a general Bayesian perspective.

I am, yes. Although as I understand it, given empirical frequencies,
conditional probability
(http://en.wikipedia.org/wiki/Conditional_probability) as deployed
here is accepted by all major statistical schools, so there needn't be
any specifically Bayesian about it. I'm still a beginner in
statistics, so I could be wrong about this.

--
gwern
http://www.gwern.net

[Aman Abdullahi Idle]

What on earth did i just read 0_o

Sent from my iPhone

--
You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.

To view this discussion on the web visit https://groups.google.com/d/msg/brain-training/-/esQG5QltckMJ.

[☉]

On Tuesday, July 31, 2012 5:09:33 PM UTC-5, EleazarWheelock wrote:

1.  The simple fact is that outside the orbit of Vulcan most arguments which are rationally compelling are not deductively valid. 

This is painfully obvious. To say it's "tantamount to obscurantism" is patently untrue. Going back to Aristotle's Rhetorics, it's well known in the philosophical literature that deductive logic is of qualified utility when trying to account for inductive inference and scientific reasoning. Recognition of this is why we have an entire discipline called Defeasible Reasoning to assist with problems not amenable to deductive analysis. If non-deductive (aka deductively invalid) modes of reasoning like "inference to the best explanation, abduction, analogical reasoning, and scientific induction" have any merit--and they manifestly do--"positive proof" should hardly be our exclusive concern. 

This is an adorable summary of a one-sided take on my terms, which any reasonable person should nevertheless be able to follow. I am of the opinion that induction in the probabilistic case is merely a disguised form of deductive reasoning (that is, in the context of mathematical procedure), so the inferences, if they take place at all, are pre-eminently syllogistic (hence, deductive), or at any rate, abductive (qua [even fuzzy] logic). On the other hand, there are varieties of induction which need not meet the bar of deductive validity, but even Bayes' theorem is a deductive result (prior to what others have said in the name of Bayesianism).

Defeasibility here is irrelevant, because there is no evidence of which to speak (so oughtn't we remain silent?). Besides, I posited a different explanation - nay, a better one ("better" meaning "more plausible") than "fraud" - viz., incompetence, but apparently you now wish to ignore this point to mischaracterize the basis of my claim, which is that a non-reply does not constitute any kind of basis on which to infer something else. The hypothesis of incompetence here reflects on the authors only when they've responded to the effect that the data has been lost, is irretrievable, etc. In no way can anyone draw a sensible conclusion from a non-response, and if anyone's degree of belief in the hypothesis of fraud on the part of a non-responder is increased thereby, then there's something going on in that believer's head in need of fixing, rather than constituting a measure of "good sense" (sensu Duhem).

 

2. Since when does Gwern have to explicitly claim he's arguing from a Bayesian perspective for me to legitimately say it is such? It's bloody obvious if you attend to the evidence. You might as well claim he must preface all cosmological claims with the gravitational constant and relevant Newtonian formulae in order to conclude he believes in gravity. What's more, he's welcome to take exception to my formalization of his reasoning. Ironically, I take the lack of evidence of his disagreement as evidence of the absence of disagreement! And even if he weren't consciously Bayesian, I can without loss of generality--and did--make an equivalent argument in Bayesian terms and thus it can rightly be termed a Bayesian perspective. Criteria 2 and 3 are also completely unnecessarily for this argument to be justifiably termed as coming from a general Bayesian perspective.

He has to do so primarily because there's more than one kind of Bayesian perspective, Allan, and within the context of statisticization (qua probability), there is more than one kind of interpretation of the same besides a Bayesian one. (There's plenty of room for argument that probabilities, if they're to be meaningful, should represent frequencies in observation, not degrees of belief.) This is called "background information", something which you apparently didn't take into account in your illogical leap to the conclusion which wasn't warranted by the total evidence of the discussion at that time - something a proper Bayesian would at least have acknowledged. Then again, you also believe that non-evidence confers (some) support to any degree of belief on some hypothesis, so it is unsurprising of you to make such an unwarranted claim.

I make no such claim about Newtonian laws, which have been thoroughly disproved anyway. A cute scarecrow, Allan, but one which doesn't scare anyone with a sense of reason rooted in fact, instead of in the presumption of mere degrees of belief. :)

 

Yes, I would never think that what you've written actually represents gwern's "reasoning" on the matter, but you seemed quite sure of this at the time. Nevertheless, if my observations are correct, I'd say you've approximated it to a fair degree. Bravo.

3. Validity here "certainly doesn't apply to anything pertaining to Bayesianism"? Wrong again. Valid means exactly what it normally means. If you grant the truth premises 1, 2, & 3, the truth of the conclusion necessarily follows. If you wish to not assent to any premises---stated or implied (like the consistency or coherency of Bayesianism, arithmetic, probability theory, etc)--that's of course your prerogative. My point was simply to give a formal and consistent account of someone's reasoning. That's all. I'm quite serenely agnostic on the ultimate truth of the argument. But my prima facie opinion is that while silence is evidence, it's very weak evidence. Where the rubber meets the road, we are in broad agreement. 

Bayesian personalism or even Bayesian inter-subjectivism (objectivism) does not have anything within the mathematics which informs us that we're to wash out today's priors and replace them with tomorrow's posteriors. The fact that you do not grasp this hints you do not understand any worthwhile meaning of validity w.r.t. Bayesianism to begin with, but even were that the case, I made it explicit what I meant by it, more in keeping with the mathematical (formal) meaning thereof. Coherence and consistence (and even the axioms of probability), as is widely known, constitute a very weak constraint on the priors within Bayesianism - such that any number of eccentricities thereby become spun by the spinneret's web of belief - and is no bar against the obvious enough objection that "belief" has no claim on the actual facts themselves.

 

4. "By the way, what you just proved (that is, that which you've put forward prior to any given refutation), as noted within the quotation marks, has nothing to do with the notion that absented evidence is positive evidence of something else (besides the absence w.r.t. the evidence)."  On one interpretation, this is exactly wrong. The proof is not tautologous. It does not say "absence of evidence is absence of evidence". But I'm not sure how you're using the term "positive evidence" or "something else". (E.g. "Absence of evidence for the Yeti is evidence for his (existential) absence." I'm afraid existential absence (nonexistence) is indeed quite "something else" than mere evidentiary absence.) Or what exactly you mean by "absence w.r.t to the evidence".  If by the latter you mean my conclusion is contingent on the premise of a defeasible claim, viz. the absence of evidence, then, yes. But that's perfectly obvious and non-fatal...Saying "absence of evidence is evidence of absence (conditioned on the absence of the evidence)" is just needlessly appending the premise to the conclusion with a pretense to insight. 

If you do not grasp "absence of evidence" never giving rise to a deducible result, then I'm afraid the conversation will have to end rather abruptly. Your confusion is most likely symptomatic of a critical fault with which I have nothing to do, but which I have taken some pains to illustrate.

5. Perhaps you'd like to give us a positive proof of the Divinity via Bayesianism as well? :)" 

I'm not even sure it's coherent to talk of "positive proof" via Bayesianism.

I hope you are sure it isn't coherent, because to speak of (one kind of) Bayesianism as conferring evidence for a hypothesis absent any evidence is about as incoherent as it gets. Not even a proper Bayesian would believe that (and if they did, they might be more... circumspect... about it).

 

You have to be careful how you deploy "evidence" "positive evidence" and "proof". Bayesian claims are always subject to defeaters. It doesn't axiomatically define, declare and deduce unassailable "positive proofs". That's much of the point I've been trying to get across...

If I'm not mistaken (about what you mean), that's also much of my own point, if you've been paying any attention at all, instead of reading in your presumptions every other post. You're something of a wearying knucklehead, I must admit, Allan. Don't worry, I get that sometimes as well.

 

That said, if a proof of the Original Mind were forthcoming, I'd hope you'd have the humility to accede regardless of its humble provenance. 

"..good tidings may yet come from [utterly worthless] Nazareth." 

Allan, it's all you now. :)

 

There is a God. And Allan Carl Jackson III is his prophet. 

If anyone wants a summary, because this is really all tortured nonsense, then I'll break it down: it is unknown whether the paper is fraudulent, but it is most likely not worthy of inclusion for evidential support of the DNB-Gf hypothesis, that is, it's real value is hitherto inconclusive and therefore null.

I'll leave the fantastical coherentism of our clownish friends to transport us whither we should otherwise go - away from reality and to the madhouse.

I hope that is understood as a sign of endearment.

argumzio 

[Aman Abdullahi Idle]

Just when i thought it could not get worse 0_0

Sent from my iPhone

--

You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.

To view this discussion on the web visit https://groups.google.com/d/msg/brain-training/-/vMJuV-AT1w8J.