Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Roberto GANAU" <roberto.ganau@studenti.unipd.it> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: fractional response model for balanced panel |
Date | Mon, 16 Jan 2012 12:05:44 +0100 |
Dear Statalist, I am working on a panel dataset of N=582 firms and T=2 years. My dependent variable is a proportion, i.e. export sales/total sales. Using Stata, I run: xtgee dep_var varlist, family(binomial) link(logit) robust Is this specification correct? Is it necessary to include time average variables and/or year dummies? Excluding the logit transformation [log(y/(1-y))], can you suggest me alternative solutions to model fractional response variables maintaining the panel structure of the data? I thank you in advance for your time and attention. Best regards, Roberto Ganau