deviation contrast for linear mixed model?
Locked
deviation contrast for linear mixed model?
![Putnick, Diane (NIH/NICHD) [E]](/images/avatar100.png "Putnick, Diane (NIH/NICHD) [E]"){kind=avatar} 
|
Hi there,
I am trying to model a repeated measures ANOVA type model with 3 repeated covariates using MIXED. Parent gender (mother vs. father) is the repeated fixed factor, country (9 groups) is the between-subjects fixed factor, and the covariates are mother-father age, education, and social desirability bias. The problem I'm having is that I want to get a deviation contrast (deviation from the grand mean) for the main effect of country. I don't see any easy way to do this in the syntax. My options seem to only include using a single group as the contrast group. Any suggestions? Thanks, Diane MIXED dv BY Parent NewCountryID /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR (0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=Parent NewCountryID Parent*NewCountryID | SSTYPE(3) /METHOD=REML /PRINT=SOLUTION TESTCOV /REPEATED=Parent | SUBJECT(idnew) COVTYPE(CSH) /EMMEANS=TABLES(Parent) COMPARE ADJ(LSD) /EMMEANS=TABLES(NewCountryID) COMPARE REFCAT(??) ADJ(LSD) /EMMEANS=TABLES(Parent*NewCountryID) . ===================== 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 |
|
|
You might be able to use the TEST subcommand.
|
Re: deviation contrast for linear mixed model?
|
|
In reply to this post by Putnick, Diane (NIH/NICHD) [E]
Thanks for the idea. I can't seem to get it to run when I have a repeated
variable in the model. I was using the following syntax (simplified by excluding the covariates for now). I keep getting the warning "Custom hypothesis test 1 (Deviation Contrasts for NewCountryID) will not be performed because the L matrix is not estimable." However, if I simplify the model, removing the repeated factor (Parent), it runs fine. So... is there any reason I can't compute the contrasts by hand? I could compute the grand mean and do a one-sample t-test for each estimated marginal mean... anything wrong with that? MIXED dv BY Parent NewCountryID /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR (0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=Parent NewCountryID Parent*NewCountryID | SSTYPE(3) /METHOD=REML /PRINT= SOLUTION TESTCOV /REPEATED=Parent | SUBJECT(IDnew) COVTYPE(CSH) /EMMEANS=TABLES(Parent) /EMMEANS=TABLES(NewCountryID) COMPARE ADJ(LSD) /EMMEANS=TABLES(Parent*NewCountryID) /TEST = 'Deviation Contrasts for NewCountryID' NewCountryID 8/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9; NewCountryID -1/9 8/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9; NewCountryID -1/9 -1/9 8/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9; NewCountryID -1/9 -1/9 -1/9 8/9 -1/9 -1/9 -1/9 -1/9 -1/9; NewCountryID -1/9 -1/9 -1/9 -1/9 8/9 -1/9 -1/9 -1/9 -1/9; NewCountryID -1/9 -1/9 -1/9 -1/9 -1/9 8/9 -1/9 -1/9 -1/9; NewCountryID -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 8/9 -1/9 -1/9; NewCountryID -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 8/9 -1/9; NewCountryID -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 -1/9 8/9. On Thu, 13 May 2010 05:05:38 -0700, rblack <[hidden email]> wrote: >You might be able to use the TEST subcommand. > > >Diane Putnick wrote: >> >> Hi there, >> >> I am trying to model a repeated measures ANOVA type model with 3 repeated >> covariates using MIXED. Parent gender (mother vs. father) is the repeated >> fixed factor, country (9 groups) is the between-subjects fixed factor, and >> the covariates are mother-father age, education, and social desirability >> bias. The problem I'm having is that I want to get a deviation contrast >> (deviation from the grand mean) for the main effect of country. I don't >> see any easy way to do this in the syntax. My options seem to only include >> using a single group as the contrast group. Any suggestions? >> >> Thanks, >> Diane >> >> MIXED dv BY Parent NewCountryID >> /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR >> (0.000000000001) HCONVERGE(0, >> ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) >> /FIXED=Parent NewCountryID Parent*NewCountryID | SSTYPE(3) >> /METHOD=REML >> /PRINT=SOLUTION TESTCOV >> /REPEATED=Parent | SUBJECT(idnew) COVTYPE(CSH) >> /EMMEANS=TABLES(Parent) COMPARE ADJ(LSD) >> /EMMEANS=TABLES(NewCountryID) COMPARE REFCAT(??) ADJ(LSD) >> /EMMEANS=TABLES(Parent*NewCountryID) . >> >> ===================== >> 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 >> >> > >-- >View this message in context: http://old.nabble.com/deviation-contrast- >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 |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |