Talk:Heckman correction
 | This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||||||||||||
| |||||||||||||||||||||
Dr. Huber's comment on this article
editDr. Huber has reviewed this Wikipedia page, and provided us with the following comments to improve its quality:
A comment to "The two-step estimator discussed above is a limited information maximum likelihood (LIML) estimator. In asymptotic theory and in finite samples as demonstrated by Monte Carlo simulations, the full information (FIML) estimator exhibits better statistical properties. However, the FIML estimator is more computationally difficult to implement.[4]"
An other disadvantage of FIML over LIML is that it imposes stronger assumptions. Specifically, the two step approach only requires the error in the selection equation to be normally distributed, NOT in the outcome equation (see for instance Wooldridge "Analysis of Cross Section and Panel Data"
It may be worth mentioning that some assumptions of the model are (at least partially) testable, e.g. the independence of the error and regressors, see Huber and Melly: "A test of the conditional independence assumption in sample selection models", Journal of Applied Econometrics (forthcoming)." or the validity of instruments used in sample selection models, see Huber and Melly "Testing exclusion restrictions and additive separability in sample selection models", Empirical Economics, 47 (1), 75-92, 2014.
Finally, the reference to propensity score matching is misleading "See also propensity score matching", as the propensity score on which matching is based in general accounts for selection based on observed characteristics, while the Heckman sample selection model tackles selection related to unobservables. However, propensity score matching can be extended to include a second propensity score as an additional regressor that is essentially a nonparametric version of the inverse Mill s ratio and thus controls for selection on unobservables (if an instrument is at hand) - this then corresponds to a nonparametric version of the Heckit model for average treatment effect evaluation, see: Huber "Treatment evaluation in the presence of sample selection", Econometric Reviews, 33 (8), 869-905, 2014.
We hope Wikipedians on this talk page can take advantage of these comments and improve the quality of the article accordingly.
Dr. Huber has published scholarly research which seems to be relevant to this Wikipedia article:
- Reference : Martin Huber & Blaise Melly, 2012. "A test of the conditional independence assumption in sample selection models," Working Papers 2012-11, Brown University, Department of Economics.
Method section
editThe page states: Since σu > 0, the coefficient on λ can only be zero if ρ=0, so testing the null that the coefficient on λ is zero is equivalent to testing for sample selectivity.
Won't the standard deviation σu require that the confidence bounds on a significance test for the above hypothesis to be invalid? I believe a test such as a t-test will only tell us the joint significance of σu and λ, but not the significance of λ itself. How do we correct for this? — Preceding unsigned comment added by Azlefty (talk • contribs) 16:49, 29 April 2016 (UTC)
Dr. Giles's comment on this article
editDr. Giles has reviewed this Wikipedia page, and provided us with the following comments to improve its quality:
"Implementations in statistics packages[edit]
R: Heckman-type procedures are available as part of the sampleSelection package.[9][10] Stata: the command heckman provides the Heckman selection model.[11]" See also[edit]
Add: EViews: the command heckit implements both the Heckman 2-step estimator and the full maximum likelihood estimator.
(The link for the EViews package is http://www.eviews.com)
We hope Wikipedians on this talk page can take advantage of these comments and improve the quality of the article accordingly.
Dr. Giles has published scholarly research which seems to be relevant to this Wikipedia article:
- Reference : Qian Chen & David E. Giles, 2009. "Finite-Sample Properties of the Maximum Likelihood Estimator for the Poisson Regression Model With Random Covariates," Econometrics Working Papers 0907, Department of Economics, University of Victoria.