Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: Mime-Version: 1.0

From   "Jun Xu" <>
Subject   RE: st: Mime-Version: 1.0
Date   Mon, 04 Apr 2005 08:31:21 -0500

in heckman models, you need to correlate errors across two equations, and because the error terms for both probit and regression are normally distributed, it's pretty easy to derive bivariate normal, and easy for maximum likelihood estimation. For logit, might not be that easy for ml. I am not sure what about two-step procedure, I think logit might work, but you need to derive mills ratio first and then stick it into regression. However, it is inefficient. For details, you need to read Madalla 1983, which I think is relatively accessible.

Jun Xu
Department of Sociology
Indiana University

From: Jacqueline Fernandez <>
Subject: st: Mime-Version: 1.0
Date: Mon, 04 Apr 2005 13:13:53 +0800

I am using STATA for my PhD research. In my work, I want to estimate the
earnings function of women but I want to correct for sample selectivity
bias using the Heckman procedure which involves estimating a labour force
participation function. The inverse Mills ratio which is derived from the
participation function and then included in the earnings function. In
estimating the labour force participation function, do I use a probit or
logit model? Is it preferable to use probit as most researches do and why
is probit prefered in the Heckman procedure?

Jacqueline Fernandez

*   For searches and help try:
Express yourself instantly with MSN Messenger! Download today - it's FREE!

* For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index