On Wed, 29 Apr 2009, yinung at Gmail wrote:
I found that gretl provides a nice "slope" feature in
However, when independent variables are discrete, gretl computes
slopes as they are continuous variables. I compare gretl's result with
other packages, such as STATA and Matlab (the Probit output produced
by Matlab are attached below)
They will automatically detect the discrete variables in Probit (as
well as in Logit) to compute the slope (also named as marginal
effects) and mark them in a footnote:
"(*) dy/dx is for discrete change of dummy variable from 0 to 1."
Do you think our gretl should do the same thing?
Yes, that makes sense. I have now revised the slope calculation
for dummy regressors in CVS.