Clarification about Clarify

anonymity2.jpgI recently switched to STATA from SPSS. The choice seemed overdetermined, not least because of the abundance of freeware add-ons for STATA (compared with the pricey programs for SPSS). For example, Clarify, developed by Michael Tomz, Jason Wittenberg and Gary King, makes it easy to estimate predicted probabilities by simulating data, a highly useful technique, especially when graphed. (I first learned about Clarify in the Martin/Epstein legal-empirical methods stats camp.) Going through this kind of work by hand is a hassle, as my co-authors and I learned when writing Docketology.

I’ve a question about the software that seemed unanswered by the documentation, and I thought there was a chance (a slim one) that it might be something our readers could answer. Ordinarily, when estimating a model that contains two or more mutually exclusive dummies, you are supposed to omit one as a comparison. Is that true when using the estsimp command in Clarify, or, because the assumption is that omitted variables are set to their mean, you should specify a value for all variables (and thus include all of the dummies in the set.)

Hope that makes sense! Additionally, if anyone else has experiences with Clarify or questions about it to share, consider this an open forum.

You may also like...

1 Response

  1. C. Zorn says:

    The short answer is “yes;” whatever your initial model looks like is what it should be in Clarify as well. Thus, if you’ve omitted one of the dummies, that will be captured by the constant term and provide the basis for comparisons.