Dear Stata professionals,
Even after reading the manual for the implementation of the selmlog command I still cannot get the results straight. I would appreciate your ideas and solutions on the following problem. The general idea is to see whether and how voluntary audit affects firms cost of debt. To check that, I need to correct for the self-selection bias of the companies in engaging in voluntary audit.
The current setting I have is:
My selection model (at first I perform Heckman) tries to describe the firms in my sample that are willing to engage in "voluntary_audit", hence the dependent variable of the selection model is a binomial (0 or 1) and I control it on various control variables. for example: v_audit = size + exports + etc.
In the outcome regression I have the cost of debt variable, which in my case is "interest rate", as dependent variable plus a set of control variables -different from the controls in the selection model.
Since the selmlog manual says that the selection regression dependent variable needs to have more than 2 options, I created a new variable which combines whether a firm can choose between engaging in voluntary audit and if it does, whether it chooses big4 auditor. So this new variable has 0,1,2 values. Then the whole setting looks like:
----- selmlog irate x1 x2 x3 x4, select(the new variable with 0,1,2 values = the controls from the selection model as I mentioned above) dfm (2) bootstrap(100)

The problem that appears is having "non-missing values" and "cannot find varlist_m". I would appreciate if someone can help me around this! Thanks