st: sample selection bias: rho and wald test

[sara.martinezdemorentin@unavarra.es]

Hi everyone,

I have a doubt related to the heckprob command. I want to estimate the
determinants of the person that carries out performance appraisal using
firm data. Since I only have observations of the dependent variable for
those firms that have actually implemented a performance appraisal system,
I want to control for potential sample selection bias. The dependent
variable is dichotomous, so I use the heckprob module of Stata. I run the
regression and get puzzling results. The estimated coefficient for rho is
not statistically significant from 0, which I think suggests that there is
not a selection bias. However, the wald test reports that I should reject
the null hypothesis that the equations are independent. The obtained rho
lies outside the -1, +1 range. Could this be the cause of the opposing
results concerning the existence of sample selection?

Thanks in advance,
Sara

*
*   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/