Binary dependent variable sem I need to test a framework and I was planing to apply SEM. 
My dependent variable is binary, as most of the explanatory (some are also categorical, and few continuous).
Witch package/tool do you recommend?
Thanks in advance for your help
 A: OpenMx is a flexible package for Structural Equation Modeling in R. It can be obtained via the website or via CRAN. Here is a link to its documentation: https://openmx.ssri.psu.edu/documentation
OpenMx allows joint models of ordinal and continuous models using threshold models for ordinal variables. Under this approach, your binary variables are assumed to follow a normal distribution that is partitioned using a threshold into binary responses.
OpenMx allows for the inclusion of continuous and ordinal variables in the same model, as well as models with only continuous or only ordinal variables. Any number of continuous variables may be included in an OpenMx model. For technical reasons (costly numerical integrations in model fitting), you should limit the number of ordinal variables to only a few. 
OpenMx's user guide has great examples on how to set up such models: https://vipbg.vcu.edu/vipbg/OpenMx2/docs//OpenMx/latest/Ordinal_Path.html
PS: As for exogenous observed variables, you may not need to model them using the ordinal approach.
A: As @Brandmaier said, OpenMx will do the trick. But almost any other SEM software (except, AFAIK, Amos) will do. I'd use Lavaan within R, but Mplus is also widely used.
You can convince yourself by running a logistic regression, then running an equivalent analysis in your SEM software of choice, and seeing if you get the same answer (you should). 
