|
|
Indirect effects with categorical med... |
|
Message/Author |
|
|
Dear Dr. Muthen, I am running a path model involving binary exogenous variables, binary and continuous mediators, and a final binary endogenous measure. I would like to test the significance of some of the indirect paths of the model, but I am writing for a field/audience more accustomed to log-odds and odds ratios than to probits. I am considering running the WLSMV analysis, explaining one indirect path in terms of the magnitude and significance of its probit coefficient, such that people can try to understand them if they want to, and then focusing solely on the significance values (ignoring the coefficients) for all other paths. At the same time, I know I can also run the analysis approximating all variables as continuous. I notice that inferences obtained this way seem to be more conservative, which might be desirable given the circumstances. I am very interested in your opinion of (1) this significance-only approach and (2) which method (probit or linear) would better approximate the results as if indirect effects could be calculated using the logit option. Any alternate suggestions you have are also most welcome. Thank you. |
|
|
I would not treat the variables as continuous if you get noticeably different results. See the Version 7.2 Language Addendum on the website with the user's guide. It describes new indirect effects using MODEL INDIRECT. |
|
Back to top |
|
|