Simon Schus posted on Wednesday, June 24, 2015 - 7:12 am
I have collected data from five groups of people and have created a SEM to estimate the relationships amongst some of the latent variables in the dataset (there was a number of missing data so I used ML).
Aside from the relationships which I have estimated without a problem I am trying to determine whether there are significant differences between the five groups for some of the latent variables.
I would usually use a one-way ANOVA and then post hoc tests to explore whether there are differences on a variable between the five groups. Is that possible when using SEM? I could not work out how to set the latent variable up as the dependent variable in an ANOVA, etc.
I assume you are doing this as a multiple group analysis using the GROUPING option. In this case, you can test differences in parameters across groups using chi-square difference testing or MODEL TEST. See pages 486-487 and 697 of the user's guide.
I am trying to run a one way ANOVA-equivalent test of differences on a manifest continuous variable between three groups. I used the following code:
USEVARIABLES ARE GROUP Y; GROUPING = GROUP (0=G1 1=G2 2=G3);
Analysis: Type=MGROUP; Estimator = ML; Model: Y(1) ; Model G1: [Y] (A1); Model G2: [Y] (A2); Model G3: [Y] (A3);
But I am not sure how to interpret the output/find the difference test results. 1) Do the model estimates and p-values represent the chi-square for each group compared to the other two groups combined? 2) Is there a specific command I can use to get this test and an output command?