Univariate Analysis
Rishi
Thanks,
Rishi
Reef Fish
I am sure some one can! :-)
It means the analysis of ONE variable.
Do you have a question about some particular univariate analysis
problem?
-- Reef Fish Bob.
P.S. On second thought, I changed my mind of "sure some one can"
because I've spend weeks not able to help m00es UNDERSTAND
(that was the keyword in your question) a simple univariate
regression problem.
ju...@microserf.org.uk
Sure.
A 'variate' is a quantity that can change---it's something you're
interested in analysing (for example, the heights of schoolchildren).
The prefix 'uni' means 'single'. 'Univariate' simply means you have a
single quantity (e.g. a scalar) that you measure and want to analyse.
So this is just the 'regular' kind of statistics that you may be taught
in school.
The word 'univariate' is typically used to make clear that
'multivariate' analysis is not the topic of interest. 'Multi' means
'many', and 'multivariate' analysis involves quantities that are
composed of several distinct quantities (e.g. a single multivariate
measurement might be the height of a schoolchild, their age, their mean
daily calorie intake and their grade in a statistics test). In
multivariate analysis, we take all of these 'measurements' as a single
measurement---a multivariate quantity (you can think of such
measurements as being vectors).
Hope that helps
C
Reef Fish
Your explanation is very good, for most readers.
Unfortunately, your LAST PARAGRAPH is the source of CONFUSION
among several of the "self-proclaimed-statisticians" such as Richard
Ulrich, who had been "teaching" others in this group for 10 YEARS
(about his own errors) but who was (and I hope is no longer) throughly
confused by the use of "multiple regression" and "multivariate
regression"
-- about the STANDARD usage in statistics.
Multiple Regression is a UNIVARIATE statistical method involving
ONE Dependent variable Y and one or more X's.
Multivariate Regression is a MULTIVARIATE statistical method
involving MORE THAN ONE Dependent variable Y in
the regression.
So, in statistics in particular, there is the need to make sure in
the explanation that "multivariate" is NOT synonymous to
"multiple" (or several variables).
Your explanation is fine for beginning stat. users who ask the question
(which the OP seems to be), because they would not have gone far
enough to be aware of the DIFFERENCE between "multiple" and
"multivariate".
But there ARE quite a few readers in this group who know just
enough to be confused by their mistaken usage of those terms.
-- Reef Fish Bob.
m00es
Reef Fish wrote:
> P.S. On second thought, I changed my mind of "sure some one can"
> because I've spend weeks not able to help m00es UNDERSTAND
> (that was the keyword in your question) a simple univariate
> regression problem.
Nice one, Reef Fish. You managed yet again to slip an insult into a
completely unrelated post.
m00es
Reef Fish
If that's an insult, that is one that is right on target, because m00es
is
STILL making the same errors in the Univariate Analysis of a
simple regression problem.
Several posts by m00es TODAY m00es was making the same blunder
in Univeriate Analysis of a regression.
Need anyone say more?
-- Reef Fish Bob.
David A. Heiser
"Rishi" <rishi....@gmail.com> wrote in message
news:1160746835.8...@b28g2000cwb.googlegroups.com...
> Can some help me understand what is Univariate Analysis.
>
> Thanks,
> Rishi
Outside of the response by junk6 and reeffish to your question, I would take
the position that it is defined more by example than by concept. Univariate
analysis usually includes: average-mean (there are many types),
variance-standard deviation, kurtosis, skewness, mode, min, max, etc. in
which the order of the values is not important. If the data is ranked by
size (and this does not change the information content, then percentiles and
related measures can be included as univariate measures. If however the
sequence of the data items is important, then the problem becomes one of
multivariate analysis, since the location of each item of data in the list
represents another variable.
The NIST set of univariate tests on computer software, however does not make
this distinction. Just like every thing else in statistics, usage is by some
local conventions.
David Heiser
Richard Ulrich
<large_nass...@yahoo.com> wrote:
"Multivariate" occasionally means "multiple dependent
variables".
If you check for STANDARD usage, say, by using
google < define:multivariate analysis >
you will find the de-facto standard is to consider
"many variables" as the meaning of multivariate.
Eleven of the twelve "Definitions ... found on the Web"
do not specify "multiple *dependent* measures."
"Multivariate multiple regression" is moderately well
established, as having several variables on each side of the
regression equation. "Multivariate regression" is usually
a mis-statement by people who mean to say multiple
regression, though Reef Fish -- and some others --
would give it the particular meaning.
The old definition which some people still respect takes
"multivariate" as referring only to the left hand side of the
equation -- if I may get rid of the fuzzy term, "dependent."
My 1973 edition of Kerlinger and Pedhazer, Multiple
Regression in Behavioral Research, has a footnote on
page 2, which includes this comment.
"We prefer to consider all analytic methods that have more
than one independent variable or more than one dependent
variable or both as multivariate methods. ... Although the point
is not all-important, it needs to be clarified early to prevent
reader confusion."
- So statisticians, even textbooks, have been abandoning the
restricted term for all of Bob's career. However, Reef Fish
Bob is particularly bad on accepting references, including
dictionaries; and his education on some vital matters of
regression (inference, multicollinearity, etc.) seems to be
dated from the 1950s, without much learning in recent decades.
(And, almost inevitably, Bob expresses nasty opinions about
whoever contradicts him. My take is, he missed out on boyhood
training in discussion and debate.)
>
> Multiple Regression is a UNIVARIATE statistical method involving
> ONE Dependent variable Y and one or more X's.
>
> Multivariate Regression is a MULTIVARIATE statistical method
> involving MORE THAN ONE Dependent variable Y in
> the regression.
>
> So, in statistics in particular, there is the need to make sure in
> the explanation that "multivariate" is NOT synonymous to
> "multiple" (or several variables).
>
> Your explanation is fine for beginning stat. users who ask the question
> (which the OP seems to be), because they would not have gone far
> enough to be aware of the DIFFERENCE between "multiple" and
> "multivariate".
>
> But there ARE quite a few readers in this group who know just
> enough to be confused by their mistaken usage of those terms.
The final sentence? -- That's possible.
--
Rich Ulrich, wpi...@pitt.edu
http://www.pitt.edu/~wpilib/index.html
Reef Fish
Multivariate regression ALWAYS means "multiple dependent variables"
in any reputable statistics textbook on Multivariate Statistical
Analysis.
>
> If you check for STANDARD usage, say, by using
> google < define:multivariate analysis >
As usual, you use the WRONG search.
Search for "Multivariate REGRESSION".
We've gone over this topic before, and you're reproducing the
same ERROR in definition and same error in Google search.
> "Multivariate multiple regression" is moderately well
The "multiple" is unnecessary. Multivariate regressioni may have
one or more indepdent variables. They are all called multivariate
regression IF AND ONLY IF there are two more more dependent
variables.
> established, as having several variables on each side of the
> regression equation. "Multivariate regression" is usually
> a mis-statement by people who mean to say multiple
> regression, though Reef Fish -- and some others --
> would give it the particular meaning.
>
> The old definition which some people still respect takes
> "multivariate" as referring only to the left hand side of the
> equation -- if I may get rid of the fuzzy term, "dependent."
>
> My 1973 edition of Kerlinger and Pedhazer, Multiple
> Regression in Behavioral Research, has a footnote on
> page 2, which includes this comment.
This is (a) a book on MULTIPLE regression, not MULTIVARIATE
regression. (b) he only said multiple regression MAY be
considered a multivariate method. That does not contradict the
fact that Multivariate REGRESSION always have two or more
dependent variables.
> "We prefer to consider all analytic methods that have more
> than one independent variable or more than one dependent
> variable or both as multivariate methods. ... Although the point
> is not all-important, it needs to be clarified early to prevent
> reader confusion."
>
> - So statisticians, even textbooks, have been abandoning the
> restricted term for all of Bob's career. However, Reef Fish
> Bob is particularly bad on accepting references, including
> dictionaries; and his education on some vital matters of
> regression (inference, multicollinearity, etc.) seems to be
> dated from the 1950s, without much learning in recent decades.
>
> (And, almost inevitably, Bob expresses nasty opinions about
> whoever contradicts him. My take is, he missed out on boyhood
> training in discussion and debate.)
That's standard Richard Ulrich ad hominem babble without showing
any evidence that a MULTIVARIATE regression is anything other
than a regression with multiple DEPENDENT variables.
Just another sign of Richard Ulrich's bankrupcy in statistical
knowledge.
>
> >
> > Multiple Regression is a UNIVARIATE statistical method involving
> > ONE Dependent variable Y and one or more X's.
I stand on that until you show ONE statistical textbook definition to
the contrary, calling that a MULTIVARIATE regression.
> >
> > Multivariate Regression is a MULTIVARIATE statistical method
> > involving MORE THAN ONE Dependent variable Y in
> > the regression.
I stand on that until you can show ONE statistical textbook
definition of Multivariate REGRESSION that says otherwise.
> >
> > So, in statistics in particular, there is the need to make sure in
> > the explanation that "multivariate" is NOT synonymous to
> > "multiple" (or several variables).
> >
> > Your explanation is fine for beginning stat. users who ask the question
> > (which the OP seems to be), because they would not have gone far
> > enough to be aware of the DIFFERENCE between "multiple" and
> > "multivariate".
Now you see why I warned the poster who signed "C" why his
explanation to the OP is not sufficient for this group.
> >
> > But there ARE quite a few readers in this group who know just
> > enough to be confused by their mistaken usage of those terms.
>
> The final sentence? -- That's possible.
Richard Ulrich has proven that he is ONE of those who is confused
by those terms, BEFORE now, and NOW. That is why Richard
Ulrich is the undisputed Chief Quack of these groups.
-- Reef Fish Bob.
m00es
Reef Fish wrote:
> m00es wrote:
> > Reef Fish wrote:
> > > P.S. On second thought, I changed my mind of "sure some one can"
> > > because I've spend weeks not able to help m00es UNDERSTAND
> > > (that was the keyword in your question) a simple univariate
> > > regression problem.
> >
> > Nice one, Reef Fish. You managed yet again to slip an insult into a
> > completely unrelated post.
> >
> > m00es
>
> If that's an insult, that is one that is right on target, because m00es
> is STILL making the same errors in the Univariate Analysis of a
> simple regression problem.
Nope, you are the one who is still making errors.
m00es
Reef Fish
Your Mouth Dancing tune is getting very OLD.
Why don't you come up with a reference from a statistical MULTIVARIATE
textbook that defined a multivariate REGRESSION as anything BUT
regression with TWO OR MORE Dependent variables.
A multiple regression, is by definition and by exclusion, a regression
with ONE Y (the DEPENDENT variable, which you used to call X),
no matter how many INDEPDENT variables Xs there are.
For someone who didn't even have the experience to know that
the dependent variable in a regression if called Y and the
independent variable called X, (with rare exceptions for exceptional
reasons) is someone IGNORANT about regression, both theoretical
AND applied.
That's m00es.
-- Reef Fish Bob.
>
> m00es
Richard Ulrich
<large_nass...@yahoo.com> wrote:
>
> Richard Ulrich wrote:
> > On 15 Oct 2006 08:29:47 -0700, "Reef Fish"
> > <large_nass...@yahoo.com> wrote:
[snip]
RU >
> > "Multivariate" occasionally means "multiple dependent
> > variables".
RF >
> Multivariate regression ALWAYS means "multiple dependent variables"
> in any reputable statistics textbook on Multivariate Statistical
> Analysis.
See Bob slip that card from his sleeve?
Or he is hallucinating the word "regression" again, as
he did in reviewing some recent posts of mine.
This is another familiar Reef Fish tactic, if it is not
a compulsion that he can't avoid. One or the other.
- He has mis-read or lost track of the argument. He cannot
recognize what I have said.
- So he starts arguing something that is not in dispute.
We agree, and Bob has no evidence *from me* (I cannot
be responsible for his hallucinations) to think otherwise -
"Multivariate regression" in a textbook would denote multiple
dependent variables, since "multiple regression" is the
accepted name for the ordinary procedure.
Bob does not mention or dispute -
The same phrase in a sci.stat.math question will be a
mis-reference to multiple regression, almost all the time.
Bob does not mention or dispute -
Most people use "multivariate" in other places to refer to
multiple variables.
[snip, lengthy re-statements and mis-reading.]
m00es
Reef Fish wrote:
> Your Mouth Dancing tune is getting very OLD.
>
> Why don't you come up with a reference from a statistical MULTIVARIATE
> textbook that defined a multivariate REGRESSION as anything BUT
> regression with TWO OR MORE Dependent variables.
>
> A multiple regression, is by definition and by exclusion, a regression
> with ONE Y (the DEPENDENT variable, which you used to call X),
> no matter how many INDEPDENT variables Xs there are.
I never actually said ANYTHING about this. I just pointed out that you
slipped an insult into a completely unrelated post.
And I agree with you that the term "multiple regression" should be
reserved for the situation where we have one dependent variable and
multiple independent variables. On the other hand, "multivariate
regression" should be used when we have more than one dependent
variable and one or more independent variables (in the latter case, we
should probably speak of "multivariate multiple regression"). So yay,
we agree on something.
Reef Fish wrote:
> For someone who didn't even have the experience to know that
> the dependent variable in a regression if called Y and the
> independent variable called X, (with rare exceptions for exceptional
> reasons) is someone IGNORANT about regression, both theoretical
> AND applied.
Whether we denote the dependent variable as X, Y, Z, or CHICKEN is
irrelevant, as long as we clearly define it that way. I always defined
it clearly in my other posts. I agree that we *typically* define Y as
the dependent variable and X as the independent one. But the letters
themselves do not carry any meaning beyond the one we assign to them.
m00es
Reef Fish
m00es wrote:
> Reef Fish wrote:
> > Your Mouth Dancing tune is getting very OLD.
> >
> > Why don't you come up with a reference from a statistical MULTIVARIATE
> > textbook that defined a multivariate REGRESSION as anything BUT
> > regression with TWO OR MORE Dependent variables.
> >
> > A multiple regression, is by definition and by exclusion, a regression
> > with ONE Y (the DEPENDENT variable, which you used to call X),
> > no matter how many INDEPDENT variables Xs there are.
>
> I never actually said ANYTHING about this. I just pointed out that you
> slipped an insult into a completely unrelated post.
SLIP in an insult? Of m00es?
You bounced into this thread not aware of even what the argument
was about -- that of MULTIVARIATE REGRESSION, and not just
multivariate methods or multivariate analysis.
You weren't even OT or IN CONTEXT, but you just like to see your
own mouth dance.
>
> And I agree with you that the term "multiple regression" should be
> reserved for the situation where we have one dependent variable and
> multiple independent variables.
Why would I need YOU (of all the Quacks) to agree with me? This is
complete STANDARD terminology in Statistics.
> On the other hand, "multivariate
> regression" should be used when we have more than one dependent
> variable and one or more independent variables (in the latter case, we
> should probably speak of "multivariate multiple regression"). So yay,
> we agree on something.
So, you finally can claim that you were correct FOR ONCE? Not even
100% correct at that.
The "multiple" is unnecessary. The only requirement in the DEFINITION
of multivariate regression is that there are MORE THAN ONE dependent
variables.
-- Reef Fish Bob.
m00es
Reef Fish wrote:
> m00es wrote:
> > reserved for the situation where we have one dependent variable and
> > multiple independent variables.
>
> Why would I need YOU (of all the Quacks) to agree with me? This is
> complete STANDARD terminology in Statistics.
Yes, I know. What's your point?
Reef Fish wrote:
> m00es wrote:
> > On the other hand, "multivariate
> > regression" should be used when we have more than one dependent
> > variable and one or more independent variables (in the latter case, we
> > should probably speak of "multivariate multiple regression"). So yay,
> > we agree on something.
>
> So, you finally can claim that you were correct FOR ONCE? Not even
> 100% correct at that.
>
> The "multiple" is unnecessary. The only requirement in the DEFINITION
> of multivariate regression is that there are MORE THAN ONE dependent
> variables.
That's what I said. However, when there are multiple dependent and
multiple independent variables, then we may want to call it
multivariate multiple regression. The "multivariate" for the fact that
there are multiple dependent variables and the "multiple" for the fact
that there are multiple independent variables.
m00es
Reef Fish
m00es wrote:
> Reef Fish wrote:
> > m00es wrote:
> > > And I agree with you that the term "multiple regression" should be
> > > reserved for the situation where we have one dependent variable and
> > > multiple independent variables.
> >
> > Why would I need YOU (of all the Quacks) to agree with me? This is
> > complete STANDARD terminology in Statistics.
>
> Yes, I know. What's your point?
My point is why do I need a Quack to agree with me when the point is
obvious because it is 100% standard usage.
>
> Reef Fish wrote:
> > m00es wrote:
> > > On the other hand, "multivariate
> > > regression" should be used when we have more than one dependent
> > > variable and one or more independent variables (in the latter case, we
> > > should probably speak of "multivariate multiple regression"). So yay,
> > > we agree on something.
> >
> > So, you finally can claim that you were correct FOR ONCE? Not even
> > 100% correct at that.
> >
> > The "multiple" is unnecessary. The only requirement in the DEFINITION
> > of multivariate regression is that there are MORE THAN ONE dependent
> > variables.
>
> That's what I said. However, when there are multiple dependent and
> multiple independent variables, then we may want to call it
> multivariate multiple regression.
You MAY want to, but no statistical textbook or any statistician ever
uses that superflous "multiple" in multivariate regression.
The "multivariate" for the fact that
> there are multiple dependent variables and the "multiple" for the fact
> that there are multiple independent variables.
>
> m00es
No, even thought it may SEEM sensible. Did you ever see any one
use the term Multivariate Simple Regression because there is one
independent variable? Of course not. You misguided pedant.
-- Reef Fish Bob.
m00es
Reef Fish wrote:
> My point is why do I need a Quack to agree with me when the point is
> obvious because it is 100% standard usage.
Yes, I know that it's standard usage. What's your point?
Reef Fish wrote:
> m00es wrote:
> The "multivariate" for the fact that
> > there are multiple dependent variables and the "multiple" for the fact
> > that there are multiple independent variables.
>
> No, even thought it may SEEM sensible. Did you ever see any one
> use the term Multivariate Simple Regression because there is one
> independent variable? Of course not. You misguided pedant.
I said that we MAY want to call it that way. It may help to clarify
that "multiple" and "multivariate" are used to denote different things.
Therefore, it's not misguided. Why are you getting so flustered?
m00es
Reef Fish
m00es wrote:
I thought you had given up your personna to Luis A. Afonso, who has
taken
up your NOISE broadcast now.
> Reef Fish wrote:
> > My point is why do I need a Quack to agree with me when the point is
> > obvious because it is 100% standard usage.
> > m00es wrote:
> > The "multivariate" for the fact that
> > > there are multiple dependent variables and the "multiple" for the fact
> > > that there are multiple independent variables.
> >
> > No, even thought it may SEEM sensible. Did you ever see any one
> > use the term Multivariate Simple Regression because there is one
> > independent variable? Of course not. You misguided pedant.
>
> I said that we MAY want to call it that way.
The beggar MAY want to ride too is wishese were horses.
>
> m00es
You ARE misguided. No statistician will EVER mistaken Multivariete
Regression to meaning anything else other than what it is.
You can take your dancing shoes OFF your MOUTH now.
-- Reef Fish Bob.
m00es
Reef Fish wrote:
> I thought you had given up your personna to Luis A. Afonso, who has
> taken up your NOISE broadcast now.
Huh?
Reef Fish wrote:
> The beggar MAY want to ride too is wishese were horses.
What?
Your posts are becoming less and less coherent.
Reef Fish wrote:
> You ARE misguided. No statistician will EVER mistaken Multivariete
> Regression to meaning anything else other than what it is.
Multivariate means that there are multiple dependent variables. That's
all that the multivariate denotes. That still leaves some ambiguity
whether there is one or more than one independent variables. By saying
multivariate simple/multiple regression, we can clarify that issue.
It's not standard usage, but it is perfectly understandable.
m00es
Reef Fish
m00es wrote:
>
> Reef Fish wrote:
> > The beggar MAY want to ride too is wishese were horses.
>
> What?
Now m00es resort to the trick of quoting OUT OF CONTEXT.
Go READ what you wrote and what I wrote BEFORE that line.
> Your posts are becoming less and less coherent.
Because m00es is getting less and less coherent even without
quoting OUT OF CONTEXT.
> Reef Fish wrote:
> > You ARE misguided. No statistician will EVER mistaken Multivariete
> > Regression to meaning anything else other than what it is.
-- Reef Fish Bob.
m00es
Reef Fish wrote:
> Now m00es resort to the trick of quoting OUT OF CONTEXT.
> Go READ what you wrote and what I wrote BEFORE that line.
Ah, I guess insults are getting boring so you now just babble on about
quoting out of context. Interesting development! This is an
entertaining evening I must say.
m00es
Reef Fish
m00es wrote:
< NOISE>
>
> m00es
m00es
m00es