Pair T-Test & Charts

[behnaz shirazi-2]

hello everyone:

Do you know how I can have a chart based on T-test result.I run:

*************T-Tests for Material Needs-************.
T-TEST
  PAIRS= RgMtrlT1  WITH lsPRFsup(PAIRED)
  /CRITERIA=CIN(.95).

T-TEST
  PAIRS= RNCMrlT1  WITH lsNONsup(PAIRED)
  /CRITERIA=CIN(.95).

T-TEST
  PAIRS= RSMTrlT1  WITH lsSUGsup(PAIRED)
  /CRITERIA=CIN(.95).

*************T-Tests for Family Social Support************.
T-TEST
  PAIRS= RgSprtT1  WITH lsPRIsup(PAIRED)
  /CRITERIA=CIN(.95).

T-TEST
  PAIRS= RNCSrtT1  WITH lsNOCsup(PAIRED)
  /CRITERIA=CIN(.95).

T-TEST
  PAIRS= RSSprtT1  WITH lsSURsup(PAIRED)
  /CRITERIA=CIN(.95).


I want to compare "mean" through spss chart.Do you have any idea or
suggestion.


Thank you
Behnaz

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

[ViAnn Beadle]

Didn't you just answered your own question--two bars, each representing mean
for each variable?

I personally would provide two overlapping histograms with some transparency
or two overlapping normal curves with some transparency.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
behnaz shirazi
Sent: Friday, June 08, 2007 10:21 AM
To: [hidden email]
Subject: Pair T-Test & Charts

hello everyone:

Do you know how I can have a chart based on T-test result.I run:

*************T-Tests for Material Needs-************.
T-TEST
  PAIRS= RgMtrlT1  WITH lsPRFsup(PAIRED)
  /CRITERIA=CIN(.95).

T-TEST
  PAIRS= RNCMrlT1  WITH lsNONsup(PAIRED)
  /CRITERIA=CIN(.95).

T-TEST
  PAIRS= RSMTrlT1  WITH lsSUGsup(PAIRED)
  /CRITERIA=CIN(.95).

*************T-Tests for Family Social Support************.
T-TEST
  PAIRS= RgSprtT1  WITH lsPRIsup(PAIRED)
  /CRITERIA=CIN(.95).

T-TEST
  PAIRS= RNCSrtT1  WITH lsNOCsup(PAIRED)
  /CRITERIA=CIN(.95).

T-TEST
  PAIRS= RSSprtT1  WITH lsSURsup(PAIRED)
  /CRITERIA=CIN(.95).


I want to compare "mean" through spss chart.Do you have any idea or
suggestion.


Thank you
Behnaz

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

[Marta García-Granero]

Hi everybody

My ADSL is back!

According to Lang & Secic (How to Report Statistics in Medicine),the
correct way of presenting paired data should preserve the pairing
information. I'm in the middle of a job with deadline june 15th, and I
can only provide a very "quick&dirty" solution using legacy graphs. I'm
sure that there is a more elegant solution for this graph using GPL,
but I don't have the time right now to work it completely. See this
example below:

DATA LIST LIST/Before After.
BEGIN DATA
 7.6 14.7
 9.9 14.1
 8.6 11.8
 9.5 16.1
 8.4 14.7
 9.2 14.1
 6.4 13.2
 9.9 14.9
 8.7 12.2
10.3 13.4
 8.3 14.0
END DATA.

* Option #1 (the best for small samples) *.
VARSTOCASES /ID = id
 /MAKE Time 'Treadmill time (min)' FROM Before After
 /INDEX = Index(Time).
GRAPH /LINE(MULTIPLE)MEAN(Time) BY Index BY id .

If sample size is big, then the line chart gets overcrouded, and this
simpler graph will be OK then:

* Option #2 (big sample sizes) *.
GRAPH /ERRORBAR( STDDEV 1 )=Before After.

BTW, thanks to all who replied to my question concerning bidimensional
hilo graphs (John, Christian, and Stephen, who sent me a great Excel
template my students appreciated a lot). I've benn so busy and, well,
incomunicated that I hadn't thanked them yet.

Regards
ADSL-happier Marta

VB> Didn't you just answered your own question--two bars, each representing mean
VB> for each variable?

VB> I personally would provide two overlapping histograms with some transparency
VB> or two overlapping normal curves with some transparency.

VB> -----Original Message-----
VB> From:  behnaz shirazi
VB> Sent: Friday, June 08, 2007 10:21 AM
VB> Subject: Pair T-Test & Charts

VB> Do you know how I can have a chart based on T-test result.I run:

VB> *************T-Tests for Material Needs-************.
VB> T-TEST
VB>   PAIRS= RgMtrlT1  WITH lsPRFsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RNCMrlT1  WITH lsNONsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RSMTrlT1  WITH lsSUGsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> *************T-Tests for Family Social Support************.
VB> T-TEST
VB>   PAIRS= RgSprtT1  WITH lsPRIsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RNCSrtT1  WITH lsNOCsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RSSprtT1  WITH lsSURsup(PAIRED)
VB>   /CRITERIA=CIN(.95).


VB> I want to compare "mean" through spss chart.Do you have any idea or
VB> suggestion.

[Cleland, Patricia (EDU)]

Marta,

Thanks for posting the code.

When I run your first option, the graph is fine except the order is AFTER BEFORE so that when I first looked at the graph, it appeared that the performance was worse AFTER the intervention.  Is there a way to get the Index to print BEFORE first, than AFTER?

Pat

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: June 12, 2007 11:24 AM
To: [hidden email]
Subject: Re: Pair T-Test & Charts

Hi everybody

My ADSL is back!

According to Lang & Secic (How to Report Statistics in Medicine),the
correct way of presenting paired data should preserve the pairing
information. I'm in the middle of a job with deadline june 15th, and I
can only provide a very "quick&dirty" solution using legacy graphs. I'm
sure that there is a more elegant solution for this graph using GPL,
but I don't have the time right now to work it completely. See this
example below:

DATA LIST LIST/Before After.
BEGIN DATA
 7.6 14.7
 9.9 14.1
 8.6 11.8
 9.5 16.1
 8.4 14.7
 9.2 14.1
 6.4 13.2
 9.9 14.9
 8.7 12.2
10.3 13.4
 8.3 14.0
END DATA.

* Option #1 (the best for small samples) *.
VARSTOCASES /ID = id
 /MAKE Time 'Treadmill time (min)' FROM Before After
 /INDEX = Index(Time).
GRAPH /LINE(MULTIPLE)MEAN(Time) BY Index BY id .

If sample size is big, then the line chart gets overcrouded, and this
simpler graph will be OK then:

* Option #2 (big sample sizes) *.
GRAPH /ERRORBAR( STDDEV 1 )=Before After.

BTW, thanks to all who replied to my question concerning bidimensional
hilo graphs (John, Christian, and Stephen, who sent me a great Excel
template my students appreciated a lot). I've benn so busy and, well,
incomunicated that I hadn't thanked them yet.

Regards
ADSL-happier Marta

VB> Didn't you just answered your own question--two bars, each representing mean
VB> for each variable?

VB> I personally would provide two overlapping histograms with some transparency
VB> or two overlapping normal curves with some transparency.

VB> -----Original Message-----
VB> From:  behnaz shirazi
VB> Sent: Friday, June 08, 2007 10:21 AM
VB> Subject: Pair T-Test & Charts

VB> Do you know how I can have a chart based on T-test result.I run:

VB> *************T-Tests for Material Needs-************.
VB> T-TEST
VB>   PAIRS= RgMtrlT1  WITH lsPRFsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RNCMrlT1  WITH lsNONsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RSMTrlT1  WITH lsSUGsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> *************T-Tests for Family Social Support************.
VB> T-TEST
VB>   PAIRS= RgSprtT1  WITH lsPRIsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RNCSrtT1  WITH lsNOCsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RSSprtT1  WITH lsSURsup(PAIRED)
VB>   /CRITERIA=CIN(.95).


VB> I want to compare "mean" through spss chart.Do you have any idea or
VB> suggestion.

[Marta García-Granero]

Hi Pat

CPE> When I run your first option, the graph is fine except the
CPE> order is AFTER BEFORE so that when I first looked at the graph,
CPE> it appeared that the performance was worse AFTER the
CPE> intervention.  Is there a way to get the Index to print BEFORE
CPE> first, than AFTER?

Try this modified version of the VARSTOCASES command:

DATA LIST LIST/Before After.
BEGIN DATA
 7.6 14.7
 9.9 14.1
 8.6 11.8
 9.5 16.1
 8.4 14.7
 9.2 14.1
 6.4 13.2
 9.9 14.9
 8.7 12.2
10.3 13.4
 8.3 14.0
END DATA.

* Option #1 (the best for small samples) *.
VARSTOCASES /ID = id
 /MAKE Time 'Treadmill time (min)' FROM Before After
 /INDEX = Index(2).
VAL LAB Index 1'Before' 2'After'.
GRAPH /LINE(MULTIPLE)MEAN(Time) BY Index BY id .

Best regards,
Marta

[ViAnn Beadle]

The order on the axis is alphabetical. If you name your variables time1 and
time2 you'll get them in the right order.

You can also edit the chart and flip the order on the Categories tab but
this is a manual process.

You can control this through syntax if you have at least Version 14 and
would like to use the GGRAPH command in place of the GRAPH command. The sort
function on the SCALE command allows you to explicitly order categorical
values.

* Chart Builder.
GGRAPH
  /GRAPHDATASET NAME="graphdataset" VARIABLES=Index Time id MISSING=LISTWISE
REPORTMISSING=NO
  /GRAPHSPEC SOURCE=INLINE.
BEGIN GPL
  SOURCE: s=userSource(id("graphdataset"))
  DATA: Index=col(source(s), name("Index"), unit.category())
  DATA: Time=col(source(s), name("Time"), unit.category())
  DATA: id=col(source(s), name("id"), unit.category())
  GUIDE: axis(dim(1), label("Index"))
  GUIDE: axis(dim(2), label("Treadmill time (min)"))
  GUIDE: legend(aesthetic(aesthetic.color.interior), label("id"))
  SCALE: cat(dim(1), sort.values("Before", "After"))
ELEMENT: line(position(Index*Time), color.interior(id), missing.wings())
END GPL.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Cleland, Patricia (EDU)
Sent: Tuesday, June 12, 2007 10:07 AM
To: [hidden email]
Subject: Re: Pair T-Test & Charts

Marta,

Thanks for posting the code.

When I run your first option, the graph is fine except the order is AFTER
BEFORE so that when I first looked at the graph, it appeared that the
performance was worse AFTER the intervention.  Is there a way to get the
Index to print BEFORE first, than AFTER?

Pat

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Marta García-Granero
Sent: June 12, 2007 11:24 AM
To: [hidden email]
Subject: Re: Pair T-Test & Charts

Hi everybody

My ADSL is back!

According to Lang & Secic (How to Report Statistics in Medicine),the
correct way of presenting paired data should preserve the pairing
information. I'm in the middle of a job with deadline june 15th, and I
can only provide a very "quick&dirty" solution using legacy graphs. I'm
sure that there is a more elegant solution for this graph using GPL,
but I don't have the time right now to work it completely. See this
example below:

DATA LIST LIST/Before After.
BEGIN DATA
 7.6 14.7
 9.9 14.1
 8.6 11.8
 9.5 16.1
 8.4 14.7
 9.2 14.1
 6.4 13.2
 9.9 14.9
 8.7 12.2
10.3 13.4
 8.3 14.0
END DATA.

* Option #1 (the best for small samples) *.
VARSTOCASES /ID = id
 /MAKE Time 'Treadmill time (min)' FROM Before After
 /INDEX = Index(Time).
GRAPH /LINE(MULTIPLE)MEAN(Time) BY Index BY id .

If sample size is big, then the line chart gets overcrouded, and this
simpler graph will be OK then:

* Option #2 (big sample sizes) *.
GRAPH /ERRORBAR( STDDEV 1 )=Before After.

BTW, thanks to all who replied to my question concerning bidimensional
hilo graphs (John, Christian, and Stephen, who sent me a great Excel
template my students appreciated a lot). I've benn so busy and, well,
incomunicated that I hadn't thanked them yet.

Regards
ADSL-happier Marta

VB> Didn't you just answered your own question--two bars, each representing
mean
VB> for each variable?

VB> I personally would provide two overlapping histograms with some
transparency
VB> or two overlapping normal curves with some transparency.

VB> -----Original Message-----
VB> From:  behnaz shirazi
VB> Sent: Friday, June 08, 2007 10:21 AM
VB> Subject: Pair T-Test & Charts

VB> Do you know how I can have a chart based on T-test result.I run:

VB> *************T-Tests for Material Needs-************.
VB> T-TEST
VB>   PAIRS= RgMtrlT1  WITH lsPRFsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RNCMrlT1  WITH lsNONsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RSMTrlT1  WITH lsSUGsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> *************T-Tests for Family Social Support************.
VB> T-TEST
VB>   PAIRS= RgSprtT1  WITH lsPRIsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RNCSrtT1  WITH lsNOCsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RSSprtT1  WITH lsSURsup(PAIRED)
VB>   /CRITERIA=CIN(.95).


VB> I want to compare "mean" through spss chart.Do you have any idea or
VB> suggestion.

[behnaz shirazi-2]

Hi Marta
Thank you very much for your useful help.But I wonder how can I have more
than one pair because I have 8 pair which I would like to show them in one
chart.
I tried to put them all but I can not have more than one Index.

Do you have any idea?


Thanks again
Behnaz

>From: Marta García-Granero              <[hidden email]>
>Reply-To: Marta García-Granero              <[hidden email]>
>To: [hidden email]
>Subject: Re: Pair T-Test & Charts
>Date: Tue, 12 Jun 2007 19:42:35 +0200
>
>Hi Pat
>
>CPE> When I run your first option, the graph is fine except the
>CPE> order is AFTER BEFORE so that when I first looked at the graph,
>CPE> it appeared that the performance was worse AFTER the
>CPE> intervention.  Is there a way to get the Index to print BEFORE
>CPE> first, than AFTER?
>
>Try this modified version of the VARSTOCASES command:
>
>DATA LIST LIST/Before After.
>BEGIN DATA
>  7.6 14.7
>  9.9 14.1
>  8.6 11.8
>  9.5 16.1
>  8.4 14.7
>  9.2 14.1
>  6.4 13.2
>  9.9 14.9
>  8.7 12.2
>10.3 13.4
>  8.3 14.0
>END DATA.
>
>* Option #1 (the best for small samples) *.
>VARSTOCASES /ID = id
>  /MAKE Time 'Treadmill time (min)' FROM Before After
>  /INDEX = Index(2).
>VAL LAB Index 1'Before' 2'After'.
>GRAPH /LINE(MULTIPLE)MEAN(Time) BY Index BY id .
>
>Best regards,
>Marta

_________________________________________________________________
Don't just search. Find. Check out the new MSN Search!
http://search.msn.click-url.com/go/onm00200636ave/direct/01/

[Marta García-Granero]

Hi

Wednesday, June 13, 2007, 9:17:38 PM, You wrote:

bs> Thank you very much for your useful help.But I wonder how can I have more
bs> than one pair because I have 8 pair which I would like to show them in one
bs> chart.  I tried to put them all but I can not have more than one Index.

This example works with two pairs of variables, and it can be easily
extended to more:

* Sample dataset, replace by your own data *.
DATA LIST LIST/Before1 After1 Before2 After2.
BEGIN DATA
  7.6 14.7 8.1 10.3
  9.9 14.1 7.1  9.2
  8.6 11.8 8.7 11.5
  9.5 16.1 5.9  8.6
  8.4 14.7 4.1  7.9
  9.2 14.1 7.4 11.1
  6.4 13.2 7.1  8.9
  9.9 14.9 8.9 11.2
  8.7 12.2 9.0 12.4
 10.3 13.4 8.3 10.7
  8.3 14.0 6.6  9.9
END DATA.

* For 8 variables, change "Variable(2)" by "Variable(8)" *.
VARSTOCASES /ID = id
 /MAKE Result FROM Before1 TO After2
 /INDEX = Variable(2) Index(2).

* For 8 variables, add labels to Variable 1'First'.... 8'Eighth' *.
VAL LABEL Index 1'Before' 2'After'/
          Variable 1'First' 2'Second'.
GRAPH
  /LINE(MULTIPLE)MEAN(Result) BY Index BY id
  /PANEL ROWVAR=Variable ROWOP=CROSS .

HTH & regards,
Marta

[Pirritano, Matthew]

Hello all,

First of all, you often see people doing a MANOVA and following it up with univariate analyses if a main effect of a between subjects factor is significant. Is this not incorrect? Doesn't the MANOVA maximize differences across a multivariate function created from the multiple dependent variables? It doesn't just tell you if there's a difference in your groups on one of the DVs? Am I correct? If so, why don't people know this? It is very frustrating. Is there a simple reference for this that I can refer people to?

Secondly, I'm looking at a paper on infertility stress. Stress is being predicted by attachment style. The authors are looking at how well the partner's attachment style predicts the individuals stress controlling for the individual's attachment style. In other words, how well does a wife's attachment style predict a husband's stress controlling for his attachment style (attachment style is measured as being one of three categories: secure, avoidant, and ambivalent). Does this kind of analysis make sense? I have been doing a lot of reading on dyadic analyses, analyzing the couple as a unit. The prefered type of analysis seems to be multilevel modeling. Does the above analysis make work, but not provide any direct information about the interactions between the effects of attachment style on husbands' and wives' stress? Is that the only difference? The absence of an interaction effect? Is there no problem with multicollinearity of predictors or the correlation of errors? Is it okay if errors are correlated between the DV and IV in ANOVA?

Any help would be much appreciated.

Thanks,
Matt

Matthew Pirritano, Ph.D.
Assistant Professor of Psychology
Smith Hall 116C
Chapman University
Department of Psychology
One University Drive
Orange, CA 92866
Telephone [phone]
FAX [phone]

________________________________

Fra: SPSSX(r) Discussion på vegne af Cleland, Patricia (EDU)
Sendt: ti 6/12/2007 9:07
Til: [hidden email]
Emne: Re: Pair T-Test & Charts



Marta,

Thanks for posting the code.

When I run your first option, the graph is fine except the order is AFTER BEFORE so that when I first looked at the graph, it appeared that the performance was worse AFTER the intervention.  Is there a way to get the Index to print BEFORE first, than AFTER?

Pat

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: June 12, 2007 11:24 AM
To: [hidden email]
Subject: Re: Pair T-Test & Charts

Hi everybody

My ADSL is back!

According to Lang & Secic (How to Report Statistics in Medicine),the
correct way of presenting paired data should preserve the pairing
information. I'm in the middle of a job with deadline june 15th, and I
can only provide a very "quick&dirty" solution using legacy graphs. I'm
sure that there is a more elegant solution for this graph using GPL,
but I don't have the time right now to work it completely. See this
example below:

DATA LIST LIST/Before After.
BEGIN DATA
 7.6 14.7
 9.9 14.1
 8.6 11.8
 9.5 16.1
 8.4 14.7
 9.2 14.1
 6.4 13.2
 9.9 14.9
 8.7 12.2
10.3 13.4
 8.3 14.0
END DATA.

* Option #1 (the best for small samples) *.
VARSTOCASES /ID = id
 /MAKE Time 'Treadmill time (min)' FROM Before After
 /INDEX = Index(Time).
GRAPH /LINE(MULTIPLE)MEAN(Time) BY Index BY id .

If sample size is big, then the line chart gets overcrouded, and this
simpler graph will be OK then:

* Option #2 (big sample sizes) *.
GRAPH /ERRORBAR( STDDEV 1 )=Before After.

BTW, thanks to all who replied to my question concerning bidimensional
hilo graphs (John, Christian, and Stephen, who sent me a great Excel
template my students appreciated a lot). I've benn so busy and, well,
incomunicated that I hadn't thanked them yet.

Regards
ADSL-happier Marta

VB> Didn't you just answered your own question--two bars, each representing mean
VB> for each variable?

VB> I personally would provide two overlapping histograms with some transparency
VB> or two overlapping normal curves with some transparency.

VB> -----Original Message-----
VB> From:  behnaz shirazi
VB> Sent: Friday, June 08, 2007 10:21 AM
VB> Subject: Pair T-Test & Charts

VB> Do you know how I can have a chart based on T-test result.I run:

VB> *************T-Tests for Material Needs-************.
VB> T-TEST
VB>   PAIRS= RgMtrlT1  WITH lsPRFsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RNCMrlT1  WITH lsNONsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RSMTrlT1  WITH lsSUGsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> *************T-Tests for Family Social Support************.
VB> T-TEST
VB>   PAIRS= RgSprtT1  WITH lsPRIsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RNCSrtT1  WITH lsNOCsup(PAIRED)
VB>   /CRITERIA=CIN(.95).

VB> T-TEST
VB>   PAIRS= RSSprtT1  WITH lsSURsup(PAIRED)
VB>   /CRITERIA=CIN(.95).


VB> I want to compare "mean" through spss chart.Do you have any idea or
VB> suggestion.

[behnaz shirazi-2]

Thank you very much Marta.Unfortunately as you said this code with a big
sample size doesn't work.I think I have to start involving with GPL.

Thanks again for your help
Behnaz

>From: Marta García-Granero              <[hidden email]>
>Reply-To: Marta García-Granero              <[hidden email]>
>To: [hidden email]
>Subject: Re: Pair T-Test & Charts
>Date: Thu, 14 Jun 2007 11:04:08 +0200
>
>Hi
>
>Wednesday, June 13, 2007, 9:17:38 PM, You wrote:
>
>bs> Thank you very much for your useful help.But I wonder how can I have
>more
>bs> than one pair because I have 8 pair which I would like to show them in
>one
>bs> chart.  I tried to put them all but I can not have more than one Index.
>
>This example works with two pairs of variables, and it can be easily
>extended to more:
>
>* Sample dataset, replace by your own data *.
>DATA LIST LIST/Before1 After1 Before2 After2.
>BEGIN DATA
>   7.6 14.7 8.1 10.3
>   9.9 14.1 7.1  9.2
>   8.6 11.8 8.7 11.5
>   9.5 16.1 5.9  8.6
>   8.4 14.7 4.1  7.9
>   9.2 14.1 7.4 11.1
>   6.4 13.2 7.1  8.9
>   9.9 14.9 8.9 11.2
>   8.7 12.2 9.0 12.4
>  10.3 13.4 8.3 10.7
>   8.3 14.0 6.6  9.9
>END DATA.
>
>* For 8 variables, change "Variable(2)" by "Variable(8)" *.
>VARSTOCASES /ID = id
>  /MAKE Result FROM Before1 TO After2
>  /INDEX = Variable(2) Index(2).
>
>* For 8 variables, add labels to Variable 1'First'.... 8'Eighth' *.
>VAL LABEL Index 1'Before' 2'After'/
>           Variable 1'First' 2'Second'.
>GRAPH
>   /LINE(MULTIPLE)MEAN(Result) BY Index BY id
>   /PANEL ROWVAR=Variable ROWOP=CROSS .
>
>HTH & regards,
>Marta

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

[Kylie Lange]

Hi Matthew,

The following references may help with your first point. My point of
view is that the choice of a MANOVA vs multiple ANOVAs should be driven
by the research questions and how they relate to the dependent
variables. In almost all situations either one method or the other will
be appropriate; but like you, I don't like to see both done. I know
there are some though that argue that it is a legitimate strategy.

I haven't looked at these references in some time, but from memory they
should be useful:

Huberty, Carl J & Morris, John D. Multivariate Analysis versus Multiple
Univariate Analyses. Psychological Bulletin, 105(20), p 302-308, 1989
Huberty, C.J. & Smith, J.D. The study of effects in MANOVA. Multivariate
Behavioral Research, 17, p 417-432, 1982.

Hope this helps.
Kylie.


On 14/06/2007 9:34 PM, Pirritano, Matthew wrote:
> Hello all,
>
> First of all, you often see people doing a MANOVA and following it up with univariate analyses if a main effect of a between subjects factor is significant. Is this not incorrect? Doesn't the MANOVA maximize differences across a multivariate function created from the multiple dependent variables? It doesn't just tell you if there's a difference in your groups on one of the DVs? Am I correct? If so, why don't people know this? It is very frustrating. Is there a simple reference for this that I can refer people to?
>
> Secondly, I'm looking at a paper on infertility stress. Stress is being predicted by attachment style. The authors are looking at how well the partner's attachment style predicts the individuals stress controlling for the individual's attachment style. In other words, how well does a wife's attachment style predict a husband's stress controlling for his attachment style (attachment style is measured as being one of three categories: secure, avoidant, and ambivalent). Does this kind of analysis make sense? I have been doing a lot of reading on dyadic analyses, analyzing the couple as a unit. The prefered type of analysis seems to be multilevel modeling. Does the above analysis make work, but not provide any direct information about the interactions between the effects of attachment style on husbands' and wives' stress? Is that the only difference? The absence of an interaction effect? Is there no problem with multicollinearity of predictors or the correlation of errors? Is !
 it!

>   okay if errors are correlated between the DV and IV in ANOVA?
>
> Any help would be much appreciated.
>
> Thanks,
> Matt
>
> Matthew Pirritano, Ph.D.
> Assistant Professor of Psychology
> Smith Hall 116C
> Chapman University
> Department of Psychology
> One University Drive
> Orange, CA 92866
> Telephone [phone]
> FAX [phone]