I am trying to build a model in SEM, where 3 categorical demographic variables (education (4 levels), age (5 levels), and gender (2 levels)) should be included as control variables. I plan to use SPSS AMOS. However, up,o,n reading some documentations of AMOS, it does not seem AMOS can handle categorical demographic variables in SEM. Is this true? If not (AMOS does do categorical control variables), can someone show me how to do it or point me to the right references/resources? If yes (AMOS does not do categorical control variables), can you please suggest alternative software (ex: does SAS do this?)? Thanks!
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
4
No, it's not true. You enter them as dummy variables, just as you would if you wanted to control for them in regression.
Unless you want to treat the variables as ordinal - in which case you have the same problem you have in regression - either treat them as (unordered) categorical, or treat them as continuous.
answered Mar 16, 2018 at 19:50
-
$\begingroup$ Thanks, Jeremy. If I dummy code the categorical variables, would it be violating some SEM assumptions (ex: normality)? If so, which estimation methods would you recommend? $\endgroup$– Yun YangCommented Mar 16, 2018 at 20:56
-
$\begingroup$ No, it won't violate them. The distributional assumptions are of the endogenous variables (variables with arrows pointing at them), not the exogenous (variables with arrows coming out). I'd recomend robust ML, which AMOS doesn't offer. I use lavaan, which has much more functionality than AMOS (at least since last time I looked at AMOS). $\endgroup$ Commented Mar 16, 2018 at 20:58
-
$\begingroup$ Thanks I will check out lavaan. One more question. Could you tell me how to test if the overall effect for each demographic variable is significant if dummy coding all of them? For example, in AMOS, how to I test if the effect of education is significant under dummy coding? $\endgroup$– Yun YangCommented Mar 18, 2018 at 10:55
-
$\begingroup$ I guess my question is can one do partial F tests in AMOS, since in regression model, if categorical variables are dummy coded, then you can use partial F tests to test the overall effect of this categorical variable, So I was wondering if you can do the same (partial F) in AMOS. Of course, if there is a better way to test the overall effect of a variable in AMOS, I would love to know. Thanks. $\endgroup$– Yun YangCommented Mar 19, 2018 at 7:56