[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
RE: st: RE: Standard error
--- Joanne Marshall wrote:
> but if i was asked to compare the Standard error of the variable wage
> (one of the independent variable) for two regression (one OLS and one
> under heckit), and found out the SE for wage is bigger in heckit, than
> OLS, what does that imply?
Typically you would compare standard errors with the estimated
regression parameter. The z-statistic is the regression parameter
divided by the standard error. To see whether that is large or small you
look at the p-value. Comparing standard errors (or p-values) across
models is not a good idea (Gelman and Stern 2006). So if you were asked
to do that than your answer should be: "it is wrong to do that"
(probably with some explanation of why that's the case). However, given
the initial confusion about terms, I would first double check if you
understood the request correctly.
> also for rho, i had an estimate for 0.38, but no chi result. what does the
> 0.38 reflect?
Are you using the -heckman- command to estimate a heckit model? If so, and
you are getting no chi2 value your model, than that probably means that you
need to correct you model. If not, you probably should switch to -heckman-.
The rho is the correlation between error terms. See for instance:
(Breen 1996) or (Long 1997).
Hope this helps,
Andrew Gelman and Hal Stern (2006) The difference between ``significant''
and ``not significant'' is not itself statistically significant. American
Statistician 60(4) pp. 328-331.
Richard Breen (1996) Regression models: Censored, sample selected, or
truncated data. Thousand Oaks: Sage.
J. Scott Long (1997) Regression models for categorical and limited dependent
variables. Thousand Oaks: Sage
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
1081 HV Amsterdam
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
* For searches and help try: