SPSSX Discussion

Mixed model question.

Classic

List

Threaded

2 messages Options

Maguin, Eugene

Mixed model question.

I have a couple of questions about a mixed model analysis setup that I can't
answer. I hope that someone can help me better understand.

Background. Participants make a rating of desire with a partner for each of
two behaviors and for each of two perspectives (self and partner). So four
ratings for each partner if both behaviors occur. Participants have multiple
partners.

I analyzed the data as

MIXED DESIRE WITH BEHAVIOR PERSPECTIVE/
FIXED=BEHAVIOR PERSPECTIVE SEXACT*PERSPECTIVE/
RANDOM=INTERCEPT | SUBJECT(ID) COVTYPE(ID)/PRINT SOLUTION.

My level 1 model is
desire^ = INT + B1*BEHAVIOR + B2*PERSPECTIVE + B3*SEXACT*PERSPECTIVE + ERR.

Somebody else analyzed the data as

MIXED DESIRE BY BEHAVIOR PERSPECTIVE/
FIXED=BEHAVIOR PERSPECTIVE SEXACT*PERSPECTIVE/
RANDOM=INTERCEPT | SUBJECT(ID) COVTYPE(ID)/PRINT SOLUTION.

What is the level 1 model here? Also, results differ and I am at a loss to
understand and explain the differences. How do I do so?

Even though I have a level 1 model that makes sense to me, it seems to me
that the data actually have a repeated measures/multivariate structure. That
is, if there were just one partner per respondent, the data could be
analyzed as either a two within factor repeated measures or as a
multivariate anova. Can this be done in Mixed? (I believe the answer to be
No.) And, if it could done is either SAS or Stata, would it be more correct
to do so?

Thanks, Gene Maguin

=====================
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

charla3@aol.com

Re: Mixed model question.

Hi Gene-

It sounds like you have a doubly multivariate design. I had a similar problem recently and was given this sage advice:

Date: Mon, 23 Feb 2009 10:01:39 -0600 Reply-To: "Swank, Paul R" <[hidden email]> Sender: "SPSSX(r) Discussion" <[hidden email]> From: "Swank, Paul R" <Paul.R.Swank@...> Subject: Re: MIXED query Comments: To: Charla Lopez <[hidden email]> In-Reply-To: <[hidden email]> Content-Type: text/plain; charset="us-ascii"

Yes you can. You construct the data set to be univariate so that each dv is on a separate line and then add a categorical variable to describe which dv it is.

Id measure/persective dv

1 x1 10 1 x2 15 2 x1 12 2 x2 17

Then add measure into the model and any interactions of measure with other effects you might want to investigate.

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

In a message dated 7/20/2009 4:25:31 P.M. Eastern Daylight Time, [hidden email] writes:

I have a couple of questions about a mixed model analysis setup that I can't
answer. I hope that someone can help me better understand.

Background. Participants make a rating of desire with a partner for each of
two behaviors and for each of two perspectives (self and partner). So four
ratings for each partner if both behaviors occur. Participants have multiple
partners.

I analyzed the data as

MIXED DESIRE WITH BEHAVIOR PERSPECTIVE/
   FIXED=BEHAVIOR PERSPECTIVE SEXACT*PERSPECTIVE/
   RANDOM=INTERCEPT | SUBJECT(ID) COVTYPE(ID)/PRINT SOLUTION.

My level 1 model is
desire^ = INT + B1*BEHAVIOR + B2*PERSPECTIVE + B3*SEXACT*PERSPECTIVE + ERR.

Somebody else analyzed the data as

MIXED DESIRE BY BEHAVIOR PERSPECTIVE/
   FIXED=BEHAVIOR PERSPECTIVE SEXACT*PERSPECTIVE/
   RANDOM=INTERCEPT | SUBJECT(ID) COVTYPE(ID)/PRINT SOLUTION.

What is the level 1 model here? Also, results differ and I am at a loss to
understand and explain the differences. How do I do so?

Even though I have a level 1 model that makes sense to me, it seems to me
that the data actually have a repeated measures/multivariate structure. That
is, if there were just one partner per respondent, the data could be
analyzed as either a two within factor repeated measures or as a
multivariate anova. Can this be done in Mixed? (I believe the answer to be
No.) And, if it could done is either SAS or Stata, would it be more correct
to do so?

Thanks, Gene Maguin

=====================
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

What's for dinner tonight? Find quick and easy dinner ideas for any occasion.