Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Maarten Buis <maartenlbuis@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: Probit/Logit - implied probability |
Date | Tue, 17 May 2011 09:56:29 +0200 |
--- Ben Hoen asked > does anyone know how to calculate the "implied probability" after a logit regression, meaning that each independent variable is kept at its mean value? Perhaps there is an after-regression tool like margins? --- Steven Samuels answered: > I assume that you don't have Stata 11 and can't use -margins-. Pre-11, -adjust- will do what you ask, but that doesn't mean its answer will be what you would expect. See http://www.stata.com/statalist/archive/2010-07/msg01596.html and the following post by Michael Norman Mitchell. The problem discussed in that post is that there is a subtle distinction between a typical predicted probability for someone within a group and the predicted probability for someone with typical values on the explanatory variables for someone within that group. If you compute what you call "implied probability" than you are computing the latter and Steve's point is that you often want the former. I agree, but there is a catch in that the former requires more from your data than the latter. The latter, predictions at typical values of the explanatory variables, requires only that the probability of being in the data does not depend on once value on the explained/left-hand-side/y-variable (*), while the former, average predicted values, requires that the probability of being in the data does not depend on both the explained/left-hand-side/y-variable and the explanatory/right-hand-side/x-variables. David Drukker pointed this out at the 2010 Italian Stata Users' meeting: page 20 of <http://www.stata.com/meeting/italy10/drukker_sug.pdf>. For a proof of the statement that the probability of being in the data only needs to be independent with respect to y see footnote 1 in Paul D. Allison (2002) "Missing Data". Quantitative Applications in the Social Sciences, 136. Thousand Oaks: Sage. I gave a more elaborate explanation of the distinction between average predicted probability and predicted probability at average values of the explanatory variables and how to compute both in Stata < 11 in: Buis, M.L. (2007) "predict and adjust with logistic regression" The Stata Journal, 7(2):221-226. http://www.stata-journal.com/article.html?article=st0127 Hope this helps, Maarten (*) As long as you decide for yourself what typical values for your explanatory variables are and thus do not take the averages in your data too strict, which often makes sense, as we often have a fair idea what those typical values are from other sources. -------------------------- Maarten L. Buis Institut fuer Soziologie Universitaet Tuebingen Wilhelmstrasse 36 72074 Tuebingen Germany http://www.maartenbuis.nl -------------------------- * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/