sem vs gsem?

[Diana Soria]

Hi everyone.
I want to use sem or gsem to evaluate the associations of different patterns of weight change 1 year after delivery and metabolic outcomes 6 years after delivery. My independent variable is categorical (4 levels). My outcome variables, total cholesterol and LDL-cholesterol, are continuous, as well as my mediating variable (BMI at 6 years). I want to adjust by some sociodemographic variables (some categorical and some continuous). My question is which type of model (sem vs gsem) is more appropriate considering that my independent variable is categorical and my dependent variable is continuous?
Thanks

[Clyde Schechter]

-gsem- allows you to have outcome variables that are of any type; for -sem- only continuous variables, strictly speaking, can be outcomes. Neither -sem- nor -gsem- restricts the types of variables used as predictor variables. So either could be used. But you mention mediation, so I would use -sem-. That's because -sem- has a postestimation command -estat teffects- that will have Stata give you separate estimates of direct, indirect, and total effects; there is no analogous command in -gsem- (although you can just calculate the results by yourself afterward.)

[Diana Soria]

Thanks for your help Clyde. I´ll try with -sem-.