one-way (x2) vs Two-way ANOVA

[AR-4]

Dear list.
I`m a little confuse.

Which one of the non parametric test would be equivalent to a parametric
two ways anova?

I've got a design with two factors (2 and 4 levels each one, so 8
different subgroups),  not heavily un-balanced.

Why I should not do two one-way anova tests instead of one two-way
anova? (one-way vs Two-way ANOVA)




Thank you in advance.
AR.

[Marta García-Granero]

Hi AR

A> Which one of the non parametric test would be equivalent to a parametric
A> two ways anova?

It is not very used, but you can find it here:

http://www.spsstools.net/Syntax/Unclassified/InteractionInFactorialDesigns.txt

A> I've got a design with two factors (2 and 4 levels each one, so 8
A> different subgroups),  not heavily un-balanced.

A> Why I should not do two one-way anova tests instead of one two-way
A> anova? (one-way vs Two-way ANOVA)

Stratifying by one factor and analyzing the effect of the second
factor only you loose a key information: interaction.


--
Regards,
Dr. Marta García-Granero,PhD           mailto:[hidden email]
Statistician

---
"It is unwise to use a statistical procedure whose use one does
not understand. SPSS syntax guide cannot supply this knowledge, and it
is certainly no substitute for the basic understanding of statistics
and statistical thinking that is essential for the wise choice of
methods and the correct interpretation of their results".

(Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)

[AR-4]

Thank you Marta, but still some confusion.

Marta García-Granero:

> Hi AR
>
> A> Which one of the non parametric test would be equivalent to a parametric
> A> two ways anova?
>
> It is not very used, but you can find it here:
>
> http://www.spsstools.net/Syntax/Unclassified/InteractionInFactorialDesigns.txt
>
> A> I've got a design with two factors (2 and 4 levels each one, so 8
> A> different subgroups),  not heavily un-balanced.
>
> A> Why I should not do two one-way anova tests instead of one two-way
> A> anova? (one-way vs Two-way ANOVA)
>
> Stratifying by one factor and analyzing the effect of the second
> factor only you loose a key information: interaction.
>

Does it means that the results of two one-way ANOVAs will be correct but
not complete? (I miss information, but I can 'be sure' of  some other info).

As far as I know 'interaction' means that the effect of one factor
differs for every level of the second factor.
If  I'm not interested in the interaction (fact1*fact2), will be the
post-hoc test of one-way ANOVA the same than the post-hoc of the two-way
anova?

Even more complicated. If parametric tests are not recommend  (because
of lack of normality or HOV), can I consider using two 'idependent'
Kruskal-Wallis tests?


Thank again.

>
> --
> Regards,
> Dr. Marta García-Granero,PhD           mailto:[hidden email]
> Statistician
>
> ---
> "It is unwise to use a statistical procedure whose use one does
> not understand. SPSS syntax guide cannot supply this knowledge, and it
> is certainly no substitute for the basic understanding of statistics
> and statistical thinking that is essential for the wise choice of
> methods and the correct interpretation of their results".
>
> (Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)
>
>

[Dale Glaser]

Hi all....quick question......I was running a zero-inflated Poisson (ZIP) Regression in Mplus and if my understanding  per the Mplus manual is correct, a Poisson regression is conducted on the nonzero counts, and a logistic regression on the zero's which "predicts the probability of being unable to assume any value except zero" (p. 26 in Mplus manual).  I decided to compare a multiple logistic regression with SPSS (coding 1 = zero's and 1 the nonzero counts) with the Mplus output, and though the same covariates are significant (and the direction of the estimates are the same across anlayses), the logits are rather different.  For instance, the logit for Mplus for one of the GPA predictors = 3.75 whereas it is 1.599 for SPSS.  So, first of all,can I assume that Mplus reports logits as the estimates for the logistic portion of ZIP?  And why do you think there is such a difference in the estimates between softwares?  At first I thought it may be due to the estimator default of
 MPlus, that being MLR which reports robust standard errors, but when I shifted the estimator to ML in Mplus it didn't make a difference in the logit.

  Any insight would be much appreciated...........thank you..

  Dale Glaser


Dale Glaser, Ph.D.
Principal--Glaser Consulting
Lecturer/Adjunct Faculty--SDSU/USD/CSUSM/AIU
President-Elect, San Diego Chapter of
American Statistical Association
3115 4th Avenue
San Diego, CA 92103
phone: [phone]
fax: [phone]
email: [hidden email]
website: www.glaserconsult.com

[Dale Glaser]

Hi all......as I just received 15.0, and printing out the new syntax for the GENLIN (generalized linear model) procedure, does anyone besides me wish SPSS would sell a syntax reference guide (that I would be more than willing to purchase) as opposed to printing out various sections (or worse yet, printing out all 2214 pages!!); unless, my information via SPSS.com is wrong, the most recent syntax manual is v 12.0, and I know much is new since then...........for instance, until today I never knew there was a '/group' function for the autorecode command as I tend to still rely on the syntax manual (I have 11.5) as opposed to point-and-click.....................so this is my plea for SPSS to consider an updated guide for purchase............or am I sounding like an antiquated relic?!!.........................

  dale glaser


Dale Glaser, Ph.D.
Principal--Glaser Consulting
Lecturer/Adjunct Faculty--SDSU/USD/CSUSM/AIU
President-Elect, San Diego Chapter of
American Statistical Association
3115 4th Avenue
San Diego, CA 92103
phone: [phone]
fax: [phone]
email: [hidden email]
website: www.glaserconsult.com

[Peck, Jon]

Well, yes, the CSR is now 2214 pages.  It would have to be printed in at least three volumes, and it would unfortunately have to be very expensive and kill a lot of trees.  But it is fully integrated into the Help system as a PDF files (Help/Command Syntax Reference) with a TOC, and it is searchable.  (Now it only kills a lot of electrons, and they get recycled.)

You also still have F1 help in the syntax window that gives you the same information for the command containing the cursor but organized with hyperlinks.  (F1 used to just give you a terse syntax chart.)

I still have my printed copy of the SPSS 12 CSR, but personally I have come to find the online version easier to use - and it is a lot easier to carry on the train!

BTW, a lot of work has gone into clarifying/rewriting sections, correcting minor errors, reorganizing some topics, and unifying into one document that covers Base and all the options together (with the option for a command identified in the command description), so we hope regular CSR users will notice the difference.

We are still printing the Data Management book, which can be downloaded as a PDF for free or purchased.  The imminent 4th edition has grown to about 550 pages, though.  It makes better reading curled up in front of a fireplace than the CSR does IMO.

-Jon Peck
(not speaking for the Publications department)

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Dale Glaser
Sent: Tuesday, December 26, 2006 4:17 PM
To: [hidden email]
Subject: [SPSSX-L] syntax reference manual for 15.0

Hi all......as I just received 15.0, and printing out the new syntax for the GENLIN (generalized linear model) procedure, does anyone besides me wish SPSS would sell a syntax reference guide (that I would be more than willing to purchase) as opposed to printing out various sections (or worse yet, printing out all 2214 pages!!); unless, my information via SPSS.com is wrong, the most recent syntax manual is v 12.0, and I know much is new since then...........for instance, until today I never knew there was a '/group' function for the autorecode command as I tend to still rely on the syntax manual (I have 11.5) as opposed to point-and-click.....................so this is my plea for SPSS to consider an updated guide for purchase............or am I sounding like an antiquated relic?!!.........................

  dale glaser


Dale Glaser, Ph.D.
Principal--Glaser Consulting
Lecturer/Adjunct Faculty--SDSU/USD/CSUSM/AIU
President-Elect, San Diego Chapter of
American Statistical Association
3115 4th Avenue
San Diego, CA 92103
phone: [phone]
fax: [phone]
email: [hidden email]
website: www.glaserconsult.com

[Dale Glaser]

thank you for your rapid response Jon ......I completely understand the economic/efficient need for having the CSR in pdf.......and I do occasionally use the help function as you allude to..but I'm still one of those old mainframe stalwarts that fervently refer to their syntax manual.........I just find it easier to have my *.sps file open and syntax manual to the side as I'm making changes to the command structure.....but I suspect I'll just have to change with the times!!!!..................thank you......dale

"Peck, Jon" <[hidden email]> wrote:  Well, yes, the CSR is now 2214 pages. It would have to be printed in at least three volumes, and it would unfortunately have to be very expensive and kill a lot of trees. But it is fully integrated into the Help system as a PDF files (Help/Command Syntax Reference) with a TOC, and it is searchable. (Now it only kills a lot of electrons, and they get recycled.)

You also still have F1 help in the syntax window that gives you the same information for the command containing the cursor but organized with hyperlinks. (F1 used to just give you a terse syntax chart.)

I still have my printed copy of the SPSS 12 CSR, but personally I have come to find the online version easier to use - and it is a lot easier to carry on the train!

BTW, a lot of work has gone into clarifying/rewriting sections, correcting minor errors, reorganizing some topics, and unifying into one document that covers Base and all the options together (with the option for a command identified in the command description), so we hope regular CSR users will notice the difference.

We are still printing the Data Management book, which can be downloaded as a PDF for free or purchased. The imminent 4th edition has grown to about 550 pages, though. It makes better reading curled up in front of a fireplace than the CSR does IMO.

-Jon Peck
(not speaking for the Publications department)

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Dale Glaser
Sent: Tuesday, December 26, 2006 4:17 PM
To: [hidden email]
Subject: [SPSSX-L] syntax reference manual for 15.0

Hi all......as I just received 15.0, and printing out the new syntax for the GENLIN (generalized linear model) procedure, does anyone besides me wish SPSS would sell a syntax reference guide (that I would be more than willing to purchase) as opposed to printing out various sections (or worse yet, printing out all 2214 pages!!); unless, my information via SPSS.com is wrong, the most recent syntax manual is v 12.0, and I know much is new since then...........for instance, until today I never knew there was a '/group' function for the autorecode command as I tend to still rely on the syntax manual (I have 11.5) as opposed to point-and-click.....................so this is my plea for SPSS to consider an updated guide for purchase............or am I sounding like an antiquated relic?!!.........................

dale glaser


Dale Glaser, Ph.D.
Principal--Glaser Consulting
Lecturer/Adjunct Faculty--SDSU/USD/CSUSM/AIU
President-Elect, San Diego Chapter of
American Statistical Association
3115 4th Avenue
San Diego, CA 92103
phone: [phone]
fax: [phone]
email: [hidden email]
website: www.glaserconsult.com



Dale Glaser, Ph.D.
Principal--Glaser Consulting
Lecturer/Adjunct Faculty--SDSU/USD/CSUSM/AIU
President-Elect, San Diego Chapter of
American Statistical Association
3115 4th Avenue
San Diego, CA 92103
phone: [phone]
fax: [phone]
email: [hidden email]
website: www.glaserconsult.com