Wednesday, October 26, 2011

Simple Logit and Probit Marginal Effects in R

Typically, economists do not care about the estimated coefficients from binary dependent variable models, but do care about their marginal effects. There does not seem to be a package in R which performs this calculation as simply as the mfx command -- which correctly handles factor variables and estimates the standard errors accurately -- as in Stata. I wanted to write a blog post containing some simple functions I have written which offer an easy solution to this issue. However, this post would have been too long. Instead I wrote a short paper which can be downloaded at the following link. Hopefully this paper will be helpful to researchers looking to implement marginal effect calculations and also to those interested in a simple basic explanation of where these figures come from when they are produced by other software packages such as Stata.

This paper outlines a simple routine to calculate the marginal effects of logit and probit regressions using the popular statistical software package R. I compare results obtained using this procedure with those produced using Stata. An extension of this routine to the generalized linear mixed effects regression is also presented.

1 comment:

Kevin Denny said...

If you are not inclined to use this approach you could use the simple scale factors derived in this little piece of mine