[George Ford]

try nrtolerance(0.00001) and see if it fails (that's the default). then try nrtolerance(0.00002) and work up until it works. If its always giving you the same results, then I think you're good. What is "very small" after all?

[Kerstin Schmidt]

I am running three different equations. For two, nrtolerance(0.00001136) and for the other one nrtolerance(0.00001375) were the smalles tolerance levels that converged. For all different nrtolerance, the coefficients and SEs are identical. Thus, it should be good (as you said in #16)

In the main analysis, would you present the results of the LPM or Biprobit, or both?