Sobel test

[Soumya Upadhyay]

I found the following example on the UCLA site one how to run a sobel test of mediation:
sgmediation science, mv(read) iv(math). I tried the same command for my own data. However, this command does not allow us to include control variables and also doesn't allow more than one independent variables at a time. Does anyone know of a way to add control variables within the Sobel test? Also, what should we look for in the output generated by the above command? Thank you!

[William Lisowski]

Reviewing the output of help sgmediation we see that the cv option (which is not discussed in the UCLA FAQ that you referred to:

http://www.ats.ucla.edu/stat/stata/faq/sgmediation.htm

allows the specification of a list of covariates. Is that what you are looking for to include control variables within the test?

Regarding your substantive questions, someone else will have to address them - I'm not familiar with the command myself. A light reading of the UCLA FAQ and the Wikipedia article on the Sobel test does not suggest that the test is designed to test the effect of a mediator on more than one independent variable at a time.

[Roman Mostazir]

As William suggested, the 'cv' option should allow you add control variables. Alternatively you can fit the model with Structrual Equation Model, Type help sem . The indirect/mediated effect is a product term (path a x path b), therefore, needs a non-linear test. Fitting after sem, use 'nlcom'. help nlcom and the standard errors will need bootstrapping to be equivalent of Sobel's test, which can be done saving the standard erros from 'rclass' in scalar after sem.