Moderation in multi-level mixed analysis

[Zdaniuk, Bozena-3]

hello, I am analyzing a longitudinal intervention study with two different intervention groups. Participants are measured on DV pre-, post- and at 6 month followup. I would like to test the following moderation model question:
 Does baseline measured (level 2) variable ModVar moderate changes in dependent variable DV from baseline to post-test to follow-up (3 time points), across both intervention groups? I am running the syntax below. Is my thinking correct that finding a significant interaction of the ModVar X Time would indicate the moderation effect?
Any pointers towards literature, websites, etc., will also be greatly appreciated.
thanks so much,
bozena

MIXED
  DV  BY group Time WITH ModVar
  /FIXED = group Time group*Time ModVar ModVar*Time | SSTYPE(3)
  /Random = intercept | SUBJECT(id) COVTYPE(ID)
  /REPEATED Time | SUBJECT(id) COVTYPE(DIAG)
  /METHOD = ML
  /PRINT = DESCRIPTIVES SOLUTION TESTCOV R
  /EMMEANS=TABLES(group) COMPARE ADJ(BONFERRONI)
  /EMMEANS=TABLES(Time) COMPARE ADJ(BONFERRONI)
  /EMMEANS=TABLES(group*Time)   COMPARE (group) ADJ(BONFERRONI)
  /EMMEANS=TABLES(group*Time)   COMPARE (Time) ADJ(BONFERRONI).
=====================
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

[Ryan]

Hi there-

Is your "level 2" moderating variable measured at the "subject(id)" level? Is it the DV measured at baseline?

By the way, the residual covariance structure specified on the REPEATED statement assumes heterogenous variances across time points, with no temporal residual correlation. The random intercept statement is mathematically equivalent to a compound symmetric specification for the residual covariance structure. Were these specifications intentional? If they were, then I would suggest you drop the random statement and replace CS with CSH on the repeated statement.

Ryan

Sent from my iPhone

> On Jun 13, 2017, at 1:58 PM, Zdaniuk, Bozena <[hidden email]> wrote:
>
> hello, I am analyzing a longitudinal intervention study with two different intervention groups. Participants are measured on DV pre-, post- and at 6 month followup. I would like to test the following moderation model question:
> Does baseline measured (level 2) variable ModVar moderate changes in dependent variable DV from baseline to post-test to follow-up (3 time points), across both intervention groups? I am running the syntax below. Is my thinking correct that finding a significant interaction of the ModVar X Time would indicate the moderation effect?
> Any pointers towards literature, websites, etc., will also be greatly appreciated.
> thanks so much,
> bozena
>
> MIXED
>  DV  BY group Time WITH ModVar
>  /FIXED = group Time group*Time ModVar ModVar*Time | SSTYPE(3)
>  /Random = intercept | SUBJECT(id) COVTYPE(ID)
>  /REPEATED Time | SUBJECT(id) COVTYPE(DIAG)
>  /METHOD = ML
>  /PRINT = DESCRIPTIVES SOLUTION TESTCOV R
>  /EMMEANS=TABLES(group) COMPARE ADJ(BONFERRONI)
>  /EMMEANS=TABLES(Time) COMPARE ADJ(BONFERRONI)
>  /EMMEANS=TABLES(group*Time)   COMPARE (group) ADJ(BONFERRONI)
>  /EMMEANS=TABLES(group*Time)   COMPARE (Time) ADJ(BONFERRONI).
> =====================
> 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