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

st: RE: Model specification problem for the Heckman twostep model

From   "Nick Cox" <[email protected]>
To   <[email protected]>
Subject   st: RE: Model specification problem for the Heckman twostep model
Date   Thu, 12 May 2005 21:09:36 +0100

I've not tried but I believe that there is much 
more that you can do than you imply. For example, 
if you use -predict- to save residuals, you can 
then fire up -qnorm- to look at normality of residuals, 
which is usually much more illuminating than some 
ritual test. Similarly, you can knit your own plots
for examining heteros{c|k}edasticity. 

Another line of attack is the package -modeldiag-, 
a ragbag of routines for examination of erratic 
error terms. This was written up in Stata Journal 
4(4) 2004. The programs can be located via 
-search- in an up-to-date Stata. I haven't checked
that these programs will work after -heckman-, 

By the way, -whitetst- is not built-in to Stata. 
It is a user-written command. This detail is naturally 
not pertinent to your main query. 

[email protected] 

Yang Li
> Our research has applied the Heckman self-selection model 
> with the twostep
> option. We have a crucial requirement to check our model 
> specifications.
> However, with the current build-in commends in STATA such as hettest,
> whitetst, rvfplot, ovtest, linktst etc, we are unable to 
> approaching the
> regular checking on homoscedasticity, normality of residuals, 
> non-linearity
> and model specifications.
> We are appreciated if you could provide any suggestions on 
> STATA commends
> can help us to cover this requirement as a necessary for our 
> research. Or
> possibly, are there any ad hoc model specifications requires 
> to be checked
> in relation to the robustness of our heckman estimation? 

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index