st: Marginal effects of interaction term following xtlogit, pa

[Erik Aadland <erikaadland@hotmail.com>]

Dear Statalist.

I have estimated a logit model on an unbalanced panel data set using xtlogit, pa.

The model is specified as follows:

xtlogit y x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 c.x12##c.x13 baseline, noconst pa eform vce(robust) 

The influence of x13 on x12 is of primary interest. x12 is a variable ranging from 0-6 where 0 is a meaningful value. x13 is ranging from 0-2 where 0 is a meaningful value. Therefore, I do not mean center the variables.

A reviewer suggested that I compute marginal effects and confidence intervals for the interaction term.
I have computed the following:

margins, dydx(x12) at(x13 = (0(1)2)) level(95)  
marginsplot,  ylin(0) 

This code produces a marginal effects plot for the interaction term with confidence intervals.

Now, the question is: Is this marginal effects plot meaningful/correct following this logit model? 

I have read many of Maarten's posts on the list regarding logit models and interactions. I have also used his Statatip 87 (exponentiating the coefficients) in my model.

I have a nagging feeling that I am missing something here.

Any input or comments are much appreciated.

Kind regards,

Erik. 		 	   		  
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/