Comparing baseline against 3 months intervention

[Manny Estrada]

Hello all,

 

I need some help to figure out the following:

 

I have a dataset of 390 participants with two time points: baseline and 3M. We did an intervention.

 

Now we need to measure the change on several behaviors, before and after intervention.

 

My dependent variable is quantitative.

 

I'm trying to implement LINEAR REGRESSION, but I don't know where to place the baseline and 3m variable to measure the change of effect.

 

I was thinking to do 2 linear regression, one for baseline and the other for 3m, then compare both. Is this a good approach?

 

Or, substract 3M from baseline and use it as independent variable. Will this approach measure the effect of our intervention?

 

Is linear regression the right approach to measure the change between baseline and 3M?

Or is there a better approach?

 

Manny Estrada

Student

[Bruce Weaver]

Manny Estrada wrote

Hello all,

I need some help to figure out the following:

I have a dataset of 390 participants with two time points: baseline and 3M. We did an intervention.

Now we need to measure the change on several behaviors, before and after intervention.

My dependent variable is quantitative.

I'm trying to implement LINEAR REGRESSION, but I don't know where to place the baseline and 3m variable to measure the change of effect.

I was thinking to do 2 linear regression, one for baseline and the other for 3m, then compare both. Is this a good approach?

Or, substract 3M from baseline and use it as independent variable. Will this approach measure the effect of our intervention?

Is linear regression the right approach to measure the change between baseline and 3M?
Or is there a better approach?

Manny Estrada
Student

You say you did an intervention.  Does that mean there are two groups of subjects, treatment and control?  

[Manny Estrada]

Yes, we did intervention, and we have control and treatment groups.

_______________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Tuesday, August 11, 2009 3:13 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:

>
> Hello all,
>
> I need some help to figure out the following:
>
> I have a dataset of 390 participants with two time points: baseline and
> 3M. We did an intervention.
>
> Now we need to measure the change on several behaviors, before and after
> intervention.
>
> My dependent variable is quantitative.
>
> I'm trying to implement LINEAR REGRESSION, but I don't know where to place
> the baseline and 3m variable to measure the change of effect.
>
> I was thinking to do 2 linear regression, one for baseline and the other
> for 3m, then compare both. Is this a good approach?
>
> Or, substract 3M from baseline and use it as independent variable. Will
> this approach measure the effect of our intervention?
>
> Is linear regression the right approach to measure the change between
> baseline and 3M?
> Or is there a better approach?
>
> Manny Estrada
> Student
>
>

You say you did an intervention.  Does that mean there are two groups of
subjects, treatment and control?



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p24923978.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Manny Estrada wrote

Yes, we did intervention, and we have control and treatment groups.

With this design, the question usually is whether the amount of change differs in the two groups.  If that is your question, the most common way to do the analysis is with a model that has:

Y  = follow-up measure
X1 = baseline measure (the covariate)
X2 = Group (with Treatment=1, Control=0)

This is the one-way ANCOVA model.  And even though you are using the follow-up score as the DV, the t-test for X2 (group) tests the null hypothesis that the amount of change is the same in the two groups.  An easy way to demonstrate this is to re-run the model with Y = 3M - baseline, but everything else the same.  The t-tests for X2 in the two models will be identical.

But note that the ANCOVA model forces the slopes (of the fitted values) for the relationship between X1 and Y to be the same for the two groups.  Because the lines for the two groups are parallel, the difference between the groups is the same for any value of the baseline score.  You should look at a scatter-plot (with different markers for the two groups) to visually assess whether it is reasonable to assume equal slopes (parallel lines).  If it appears that the slopes are not equal, you may want to run a slightly more complicated model that also includes the X1*X2 product term (i.e., the group x covariate interaction).  The inclusion of that product term allows the lines to depart from parallel, and if the t-test on the product term is statistically significant, the difference between the groups depends on what value of baseline score you look at.  

[Manny Estrada]

Hi Bruce,

Thank you very much for answering my question.

I run your syntax, my results are attached in a .htm file.


UNIANOVA change BY grp WITH pre /METHOD = SSTYPE(3) /INTERCEPT = INCLUDE /CRITERIA = ALPHA(.05) /print = parameter /emmeans = table(grp) /DESIGN = pre grp .

UNIANOVA alcohol_change BY bhqgid WITH  ahr174c /METHOD = SSTYPE(3) /INTERCEPT = INCLUDE /CRITERIA = ALPHA(.05) /print = parameter /emmeans = table(bhqgid) /DESIGN = ahr174c bhqgid.


Now, could you please help me to interpret the results?

Thanks,

Manny Estrada
Student

________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Tuesday, August 11, 2009 5:50 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:
>
> Yes, we did intervention, and we have control and treatment groups.
>
>

With this design, the question usually is whether the amount of change
differs in the two groups.  If that is your question, the most common way to
do the analysis is with a model that has:

Y  = follow-up measure
X1 = baseline measure (the covariate)
X2 = Group (with Treatment=1, Control=0)

This is the one-way ANCOVA model.  And even though you are using the
follow-up score as the DV, the t-test for X2 (group) tests the null
hypothesis that the amount of change is the same in the two groups.  An easy
way to demonstrate this is to re-run the model with Y = 3M - baseline, but
everything else the same.  The t-tests for X2 in the two models will be
identical.

But note that the ANCOVA model forces the slopes (of the fitted values) for
the relationship between X1 and Y to be the same for the two groups.
Because the lines for the two groups are parallel, the difference between
the groups is the same for any value of the baseline score.  You should look
at a scatter-plot (with different markers for the two groups) to visually
assess whether it is reasonable to assume equal slopes (parallel lines).  If
it appears that the slopes are not equal, you may want to run a slightly
more complicated model that also includes the X1*X2 product term (i.e., the
group x covariate interaction).  The inclusion of that product term allows
the lines to depart from parallel, and if the t-test on the product term is
statistically significant, the difference between the groups depends on what
value of baseline score you look at.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p24926416.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

My apologies if this appears twice.  I changed my posting address in Nabble (in an effort to avoid all of the "out of office" messages being discussed in another thread), but forgot to subscribe to the list with the new posting address before replying to Manny.

Manny Estrada wrote

Hi Bruce,

Thank you very much for answering my question.

I run your syntax, my results are attached in a .htm file.


UNIANOVA change BY grp WITH pre /METHOD = SSTYPE(3) /INTERCEPT = INCLUDE /CRITERIA = ALPHA(.05) /print = parameter /emmeans = table(grp) /DESIGN = pre grp .

UNIANOVA alcohol_change BY bhqgid WITH  ahr174c /METHOD = SSTYPE(3) /INTERCEPT = INCLUDE /CRITERIA = ALPHA(.05) /print = parameter /emmeans = table(bhqgid) /DESIGN = ahr174c bhqgid.


Now, could you please help me to interpret the results?

Thanks,

Manny Estrada
Student

Your scatter-plot was not visible in the Nabble archive, so I can't judge how well the assumption of parallel lines fits your data.  But the F-test for bhqgid (your Group variable) is clearly non-significant (F < 1, p = .841), so you would not reject the null hypothesis (which states that the amount of change is the same in the two groups).

Also, remember that the more conventional analysis would use the 3M measure rather than the change score as the dependent variable.  The reason I ran both in that sample file you found was to show that the F-test for the group variable is the same in both models.  

[Manny Estrada]

Bruce,

I think this time you should be able to see my scatter-plot attached.
I guess my control group is doing better than my intervention.

Yes, I'm using my post variable as dependent variable instead of the change.

Also, I included Levene's Test, and interaction between my pre and group.

Your input and comments are greatly appreciated.

Regards,

Manny


________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Wednesday, August 12, 2009 2:18 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:

>
> Hi Bruce,
>
> Thank you very much for answering my question.
>
> I run your syntax, my results are attached in a .htm file.
>
>
> UNIANOVA change BY grp WITH pre /METHOD = SSTYPE(3) /INTERCEPT = INCLUDE
> /CRITERIA = ALPHA(.05) /print = parameter /emmeans = table(grp) /DESIGN =
> pre grp .
>
> UNIANOVA alcohol_change BY bhqgid WITH  ahr174c /METHOD = SSTYPE(3)
> /INTERCEPT = INCLUDE /CRITERIA = ALPHA(.05) /print = parameter /emmeans =
> table(bhqgid) /DESIGN = ahr174c bhqgid.
>
>
> Now, could you please help me to interpret the results?
>
> Thanks,
>
> Manny Estrada
> Student
>
>
>

Your scatter-plot was not visible in the Nabble archive, so I can't judge
how well the assumption of parallel lines fits your data.  But the F-test
for bhqgid (your Group variable) is clearly non-significant (F < 1, p =
.841), so you would not reject the null hypothesis (which states that the
amount of change is the same in the two groups).

Also, remember that the more conventional analysis would use the 3M measure
rather than the change score as the dependent variable.  The reason I ran
both in that sample file you found was to show that the F-test for the group
variable is the same in both models.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

NOTE:  My hotmail address is for posting only, and messages sent to it will
be deleted.

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p24942044.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Swank, Paul R]

Have you looked the the distribution of the residuals here? This data is badly skewed.

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Manny Estrada
Sent: Thursday, August 13, 2009 10:06 AM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Bruce,

I think this time you should be able to see my scatter-plot attached.
I guess my control group is doing better than my intervention.

Yes, I'm using my post variable as dependent variable instead of the change.

Also, I included Levene's Test, and interaction between my pre and group.

Your input and comments are greatly appreciated.

Regards,

Manny


________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Wednesday, August 12, 2009 2:18 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:

>
> Hi Bruce,
>
> Thank you very much for answering my question.
>
> I run your syntax, my results are attached in a .htm file.
>
>
> UNIANOVA change BY grp WITH pre /METHOD = SSTYPE(3) /INTERCEPT =
> INCLUDE /CRITERIA = ALPHA(.05) /print = parameter /emmeans =
> table(grp) /DESIGN = pre grp .
>
> UNIANOVA alcohol_change BY bhqgid WITH  ahr174c /METHOD = SSTYPE(3)
> /INTERCEPT = INCLUDE /CRITERIA = ALPHA(.05) /print = parameter
> /emmeans =
> table(bhqgid) /DESIGN = ahr174c bhqgid.
>
>
> Now, could you please help me to interpret the results?
>
> Thanks,
>
> Manny Estrada
> Student
>
>
>

Your scatter-plot was not visible in the Nabble archive, so I can't judge how well the assumption of parallel lines fits your data.  But the F-test for bhqgid (your Group variable) is clearly non-significant (F < 1, p = .841), so you would not reject the null hypothesis (which states that the amount of change is the same in the two groups).

Also, remember that the more conventional analysis would use the 3M measure rather than the change score as the dependent variable.  The reason I ran both in that sample file you found was to show that the F-test for the group variable is the same in both models.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

NOTE:  My hotmail address is for posting only, and messages sent to it will be deleted.

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p24942044.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Manny Estrada]

I transformed using log 10.
Please take a look of my attachment.

________________________________________
From: Swank, Paul R [[hidden email]]
Sent: Thursday, August 13, 2009 1:12 PM
To: Manny Estrada; [hidden email]
Subject: RE: Comparing baseline against 3 months intervention

Have you looked the the distribution of the residuals here? This data is badly skewed.

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Manny Estrada
Sent: Thursday, August 13, 2009 10:06 AM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Bruce,

I think this time you should be able to see my scatter-plot attached.
I guess my control group is doing better than my intervention.

Yes, I'm using my post variable as dependent variable instead of the change.

Also, I included Levene's Test, and interaction between my pre and group.

Your input and comments are greatly appreciated.

Regards,

Manny


________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Wednesday, August 12, 2009 2:18 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:

>
> Hi Bruce,
>
> Thank you very much for answering my question.
>
> I run your syntax, my results are attached in a .htm file.
>
>
> UNIANOVA change BY grp WITH pre /METHOD = SSTYPE(3) /INTERCEPT =
> INCLUDE /CRITERIA = ALPHA(.05) /print = parameter /emmeans =
> table(grp) /DESIGN = pre grp .
>
> UNIANOVA alcohol_change BY bhqgid WITH  ahr174c /METHOD = SSTYPE(3)
> /INTERCEPT = INCLUDE /CRITERIA = ALPHA(.05) /print = parameter
> /emmeans =
> table(bhqgid) /DESIGN = ahr174c bhqgid.
>
>
> Now, could you please help me to interpret the results?
>
> Thanks,
>
> Manny Estrada
> Student
>
>
>

Your scatter-plot was not visible in the Nabble archive, so I can't judge how well the assumption of parallel lines fits your data.  But the F-test for bhqgid (your Group variable) is clearly non-significant (F < 1, p = .841), so you would not reject the null hypothesis (which states that the amount of change is the same in the two groups).

Also, remember that the more conventional analysis would use the 3M measure rather than the change score as the dependent variable.  The reason I ran both in that sample file you found was to show that the F-test for the group variable is the same in both models.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

NOTE:  My hotmail address is for posting only, and messages sent to it will be deleted.

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p24942044.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[Swank, Paul R]

Can't see the graphs but the means and sd look better.

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston


-----Original Message-----
From: Manny Estrada [mailto:[hidden email]]
Sent: Thursday, August 13, 2009 1:11 PM
To: Swank, Paul R; [hidden email]
Subject: RE: Comparing baseline against 3 months intervention

I transformed using log 10.
Please take a look of my attachment.

________________________________________
From: Swank, Paul R [[hidden email]]
Sent: Thursday, August 13, 2009 1:12 PM
To: Manny Estrada; [hidden email]
Subject: RE: Comparing baseline against 3 months intervention

Have you looked the the distribution of the residuals here? This data is badly skewed.

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Manny Estrada
Sent: Thursday, August 13, 2009 10:06 AM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Bruce,

I think this time you should be able to see my scatter-plot attached.
I guess my control group is doing better than my intervention.

Yes, I'm using my post variable as dependent variable instead of the change.

Also, I included Levene's Test, and interaction between my pre and group.

Your input and comments are greatly appreciated.

Regards,

Manny


________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Wednesday, August 12, 2009 2:18 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:

>
> Hi Bruce,
>
> Thank you very much for answering my question.
>
> I run your syntax, my results are attached in a .htm file.
>
>
> UNIANOVA change BY grp WITH pre /METHOD = SSTYPE(3) /INTERCEPT =
> INCLUDE /CRITERIA = ALPHA(.05) /print = parameter /emmeans =
> table(grp) /DESIGN = pre grp .
>
> UNIANOVA alcohol_change BY bhqgid WITH  ahr174c /METHOD = SSTYPE(3)
> /INTERCEPT = INCLUDE /CRITERIA = ALPHA(.05) /print = parameter
> /emmeans =
> table(bhqgid) /DESIGN = ahr174c bhqgid.
>
>
> Now, could you please help me to interpret the results?
>
> Thanks,
>
> Manny Estrada
> Student
>
>
>

Your scatter-plot was not visible in the Nabble archive, so I can't judge how well the assumption of parallel lines fits your data.  But the F-test for bhqgid (your Group variable) is clearly non-significant (F < 1, p = .841), so you would not reject the null hypothesis (which states that the amount of change is the same in the two groups).

Also, remember that the more conventional analysis would use the 3M measure rather than the change score as the dependent variable.  The reason I ran both in that sample file you found was to show that the F-test for the group variable is the same in both models.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

NOTE:  My hotmail address is for posting only, and messages sent to it will be deleted.

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p24942044.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Victor Kogler]

Hi all,

I recently had to reinstall Windows and all my software (may Microsoft burn in the 7 hells) and I find I am unable to install Answer Tree 3.1.  I get the following error message: "This system does not meet an installation requirement: 32MB RAM (32,768kb)." Since I have 4GB of memory on board, I do not think this is the true problem. 

I originally installed Answer Tree under the original version Window XP and it worked fine as the operating system evolved under all the subsequent service packs.  My guess is that installing Answer Tree on top of Windows XP SP3 is somehow connected to the problem.  I realize too late that I probably should have installed Answer Tree before I started the Windows Update process.  After wasting two weekends re-installing software, I am reluctant to court disaster by rolling back the SP3 update and then trying to install Answer Tree.

Has anyone out there dealt with this problem?  Is there perhaps an Answer Tree command line installation switch that gets around the false memory error?

Thanks,


Victor Kogler

[Bruce Weaver]

Nabble told me this was not accepted by the mailing list first time around.  Apologies if it appears twice.

Manny Estrada wrote

Bruce,

I think this time you should be able to see my scatter-plot attached.
I guess my control group is doing better than my intervention.

Yes, I'm using my post variable as dependent variable instead of the change.

Also, I included Levene's Test, and interaction between my pre and group.

Your input and comments are greatly appreciated.

Regards,

Manny

The graph is not visible in the Nabble archive where I read the list.  (Folks who read the list via e-mail may be able to see it.)  The F-test on the interaction has a p-value well above .05, which suggests that nothing useful is gained by including the Group x Covariate interaction term in the model.  In other words, it suggests that the simpler ANCOVA model, which forces the lines to be parallel, is probably OK.  

Why do you say your control group is doing better?  The F-test for your group variable is not close to statistical significance (p = .388), nor is the Group x Covariate interaction (p = .304).  Only the linear relationship between covariate and dependent variable is significant (p = .011).

[SPSS Support]

Hello Victor,
   This AnswerTree installation problem occurs with computers that have 3 GB or more RAM. I've pasted a resolution below my signature to address the problem. There is a link there to an ftp site with a file to download (setup.inx) before the AnswerTree installation. I used this resolution and file to install AnswerTree on my current computer. Installing under XP SP3 has not been a problem.

David Matheson
SPSS Statistical Support

*******************************



Resolution number: 22061  Created on: Dec 3 2001  Last Reviewed on: Apr 21 2009
Resolution Status: Published - External
Problem Subject:  AnswerTree reports not enough RAM when installing.
Problem Description:  I am trying to install AnswerTree 3.1 on to my computer. During the installation, I receive an error that my server does not meet the minimum RAM requirement of 32MB. My computer has more than 4 GB of RAM on it. How can I get AnswerTree to recognize the RAM and install successfully?



Resolution Subject: The problem is only related to computers with 3 GB of RAM or more.
Resolution Description:
The problem is only related to computers with 3 GB of RAM or more. If your computer does have this amount or more of RAM, please log into the SPSS FTP site. You can do this by typing the following address in a Web browser: ftp://ftp.spss.com/pub/web/at31. Download the file setup.inx to your local hard drive. If you are using FTP software to do the file transfer instead, you will need to type anonymous for the user name and an email address for the password. Make sure you choose Binary as your format of transfer. The setup.inx file instructs the install to simply report problems as found, but not abort the install for supposed deficiencies in RAM or Virtual Memory. Copy the contents of the AnswerTree folder from the CD-ROM to a folder on the hard drive. The folder name is not important. Next copy the downloaded file setup.inx to the same location so that it replaces the setup.inx in the folder where you copied the contents of the AnswerTree folder. Run setup.exe from the copi!
 ed folder
and the error may re-appear on installation but click OK and the install will continue.


________________________________________
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Victor Kogler
Sent: Thursday, August 13, 2009 1:58 PM
To: [hidden email]
Subject: Answer Tree Installation

Hi all,

I recently had to reinstall Windows and all my software (may Microsoft burn in the 7 hells) and I find I am unable to install Answer Tree 3.1.  I get the following error message: "This system does not meet an installation requirement: 32MB RAM (32,768kb)." Since I have 4GB of memory on board, I do not think this is the true problem.

I originally installed Answer Tree under the original version Window XP and it worked fine as the operating system evolved under all the subsequent service packs.  My guess is that installing Answer Tree on top of Windows XP SP3 is somehow connected to the problem.  I realize too late that I probably should have installed Answer Tree before I started the Windows Update process.  After wasting two weekends re-installing software, I am reluctant to court disaster by rolling back the SP3 update and then trying to install Answer Tree.

Has anyone out there dealt with this problem?  Is there perhaps an Answer Tree command line installation switch that gets around the false memory error?

Thanks,


Victor Kogler

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Manny Estrada wrote

I transformed using log 10.
Please take a look of my attachment.

Taking the square root is often recommended for count data (e.g., see Dave Howell's book, Statistical Methods for Psychology).

[Manny Estrada]

Bruce,

What will be the difference of doing ONE-WAY ANCOVA vs ANCOVA REPEATED MEASURES?
Thank you for your help.

Manny

________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Tuesday, August 11, 2009 5:50 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:
>
> Yes, we did intervention, and we have control and treatment groups.
>
>

With this design, the question usually is whether the amount of change
differs in the two groups.  If that is your question, the most common way to
do the analysis is with a model that has:

Y  = follow-up measure
X1 = baseline measure (the covariate)
X2 = Group (with Treatment=1, Control=0)

This is the one-way ANCOVA model.  And even though you are using the
follow-up score as the DV, the t-test for X2 (group) tests the null
hypothesis that the amount of change is the same in the two groups.  An easy
way to demonstrate this is to re-run the model with Y = 3M - baseline, but
everything else the same.  The t-tests for X2 in the two models will be
identical.

But note that the ANCOVA model forces the slopes (of the fitted values) for
the relationship between X1 and Y to be the same for the two groups.
Because the lines for the two groups are parallel, the difference between
the groups is the same for any value of the baseline score.  You should look
at a scatter-plot (with different markers for the two groups) to visually
assess whether it is reasonable to assume equal slopes (parallel lines).  If
it appears that the slopes are not equal, you may want to run a slightly
more complicated model that also includes the X1*X2 product term (i.e., the
group x covariate interaction).  The inclusion of that product term allows
the lines to depart from parallel, and if the t-test on the product term is
statistically significant, the difference between the groups depends on what
value of baseline score you look at.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p24926416.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Manny Estrada wrote

Bruce,

What will be the difference of doing ONE-WAY ANCOVA vs ANCOVA REPEATED MEASURES?
Thank you for your help.

Manny

Do you mean one-way ANCOVA vs ANOVA (not ANCOVA) with repeated measures?  If we're still talking about the same data, you had pre- and post measures for two groups of subjects.  Sometimes, people do use a 2x2 mixed design ANOVA, with Group as a between-Ss factor and Time (pre, post) as a within-Ss factor.  The F-test for the interaction from that ANOVA model is equivalent to an independent groups t-test on the change scores.  But most authors argue that the one-way ANCOVA is a better way to go with that design.

[Manny Estrada]

Bruce

"Sometimes, people do use a 2x2 mixed design ANOVA,
with Group as a between-Ss factor and Time (pre, post) as a within-Ss
factor."

If we dont want to include group (control-experimental) could I use race instead?

Thank you for helping,

Manny

________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Tuesday, August 18, 2009 1:58 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:

>
> Bruce,
>
> What will be the difference of doing ONE-WAY ANCOVA vs ANCOVA REPEATED
> MEASURES?
> Thank you for your help.
>
> Manny
>
>

Do you mean one-way ANCOVA vs ANOVA (not ANCOVA) with repeated measures?  If
we're still talking about the same data, you had pre- and post measures for
two groups of subjects.  Sometimes, people do use a 2x2 mixed design ANOVA,
with Group as a between-Ss factor and Time (pre, post) as a within-Ss
factor.  The F-test for the interaction from that ANOVA model is equivalent
to an independent groups t-test on the change scores.  But most authors
argue that the one-way ANCOVA is a better way to go with that design.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

NOTE:  My hotmail address is for posting only, and messages sent to it will
be deleted.

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p25030319.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Manny Estrada wrote

Bruce

"Sometimes, people do use a 2x2 mixed design ANOVA,
with Group as a between-Ss factor and Time (pre, post) as a within-Ss
factor."

If we dont want to include group (control-experimental) could I use race instead?

Thank you for helping,

Manny

Yes, of course you can.  But would you not want both variables (and perhaps their interaction) in your model?  

Also note that with a variable like race (or sex, etc), the groups are less likely to be equivalent on the baseline score than groups that are formed by random assignment.  And (near) equivalence of groups on the baseline score is a very good thing when it comes to the pre-post type of design.

[Swank, Paul R]

If all you want to do is determine whether the scores taken at time 1 differ from the scores at time 2, you don't need a group. But ANCOVA is designed to look at group differences controlling for one or more covariate. Repeated measures ANOVA without a group will tell you if the means differ over time. If you have a grouping variable, ANOVA will generally be more powerful unless the correlation of pre and posttest is high (.85 or .9). However, it is hard to see how you have a pretest and posttest without an intervention. Without the intervention it is an observational longitudinal design.

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Manny Estrada
Sent: Tuesday, August 18, 2009 3:57 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Bruce

"Sometimes, people do use a 2x2 mixed design ANOVA,
with Group as a between-Ss factor and Time (pre, post) as a within-Ss
factor."

If we dont want to include group (control-experimental) could I use race instead?

Thank you for helping,

Manny

________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Tuesday, August 18, 2009 1:58 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:

>
> Bruce,
>
> What will be the difference of doing ONE-WAY ANCOVA vs ANCOVA REPEATED
> MEASURES?
> Thank you for your help.
>
> Manny
>
>

Do you mean one-way ANCOVA vs ANOVA (not ANCOVA) with repeated measures?  If
we're still talking about the same data, you had pre- and post measures for
two groups of subjects.  Sometimes, people do use a 2x2 mixed design ANOVA,
with Group as a between-Ss factor and Time (pre, post) as a within-Ss
factor.  The F-test for the interaction from that ANOVA model is equivalent
to an independent groups t-test on the change scores.  But most authors
argue that the one-way ANCOVA is a better way to go with that design.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

NOTE:  My hotmail address is for posting only, and messages sent to it will
be deleted.

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p25030319.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Manny Estrada]

Bruce, Paul and Art,

 

Thank you for answering my messages

 

How about a T Paired Test splitting by race?

 

Would it be a proper test analysis?

 

For now, I just need to compare two groups over 2 period of times.

 

Manny

---

From: Art Kendall [[hidden email]]
Sent: Tuesday, August 18, 2009 6:07 PM
To: Manny Estrada
Cc: [hidden email]
Subject: Re: [SPSSX-L] Comparing baseline against 3 months intervention

Why do separate analyses? Why not do a group by race by time 3-way ANOVA or a group by race 2-way ANCOVA.

That way you minimize the error variance used in the tests.   If you do the 3-way ANOVA the first graph you use to examine your results would have
the DV on the Y axis, time on the X axis and separate lines each race by group combination.
If you find that race is not significant the graph would only have 2 lines. 

Art Kendall
Social Research Consultants

Manny Estrada wrote:

Bruce

"Sometimes, people do use a 2x2 mixed design ANOVA,
with Group as a between-Ss factor and Time (pre, post) as a within-Ss
factor."

If we dont want to include group (control-experimental) could I use race instead?

Thank you for helping,

Manny

________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Bruce Weaver [[hidden email]]
Sent: Tuesday, August 18, 2009 1:58 PM
To: [hidden email]
Subject: Re: Comparing baseline against 3 months intervention

Manny Estrada wrote:
  

Bruce,

What will be the difference of doing ONE-WAY ANCOVA vs ANCOVA REPEATED
MEASURES?
Thank you for your help.

Manny


    

Do you mean one-way ANCOVA vs ANOVA (not ANCOVA) with repeated measures?  If
we're still talking about the same data, you had pre- and post measures for
two groups of subjects.  Sometimes, people do use a 2x2 mixed design ANOVA,
with Group as a between-Ss factor and Time (pre, post) as a within-Ss
factor.  The F-test for the interaction from that ANOVA model is equivalent
to an independent groups t-test on the change scores.  But most authors
argue that the one-way ANCOVA is a better way to go with that design.



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

NOTE:  My hotmail address is for posting only, and messages sent to it will
be deleted.

--
View this message in context: http://www.nabble.com/Comparing-baseline-against-3-months-intervention-tp24920934p25030319.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD