|
|
Conditional Multivariate Change |
|
Message/Author |
|
patelrj posted on Monday, December 01, 2014 - 5:13 am
|
|
|
Hi, I've been running analyses for a project in which pretests and posttests were administred on three dependent variables. There were two conditions, ( 0 = control; 1 = treatment). Several colleagues and I can't seem to agree on how to model our data. We want to test differences and mediation of two dependent measures on the third dependent measure. I'm hoping someone can help in settling the argument. The fit measures are very low for the model, but I am afraid that the model is misspecified rather than that the hypotheses themselves are wrong. Conceptually we've modeled the data accordingly: #Effect of treatment Treatment ~ interceptA + changeA + InterceptB + changeB + InterceptC + changeC #Dependent measures (3) difference scores InterceptA =~ 1.0*observedPretestA + 1.0*observedPosttestA ChangeA =~ 1.0* observedPosttestA InterceptB =~ 1.0*observedPretestB + 1.0*observedPosttestB ChangeB =~ 1.0* observedPosttestB InterceptC =~ 1.0*observedPretestC + 1.0*observedPosttestC ChangeC =~ 1.0* observedPosttestC #Errors of posttest observed variables Errors observedPosttestA= observedPosttestB= observedPosttestC = 0 #mediations ChangeA ~ ChangeB + ChangeC #Also indirect effects Indirect effects of Treatment-ChangeB /ChangeC-> ChangeA |
|
patelrj posted on Monday, December 01, 2014 - 5:25 am
|
|
|
Sorry, I misspecified my model in the above example: I wrote here that the dependent measures were regressed on Treatment (independent). In our model treatment was de dichotomous predictor of all the dependent measures. My apologies for the confusion |
|
|
You should post this on a general discussion forum like SEMNET. |
|
Back to top |
|
|