re: st: When to use ML or bootrapping

[Christopher Baum <kit.baum@bc.edu>]

<>
Amy said

I read the preface to Maximum Likelihood Estimation with Stata, Fourth Edition,
 http://www.stata.com/bookstore/pdf/ml4-preface.pdf

I don't see a chapter on when ML is appropriate, unless that is covered in the practical implications discussion.

As for bootstrapping, Chapter 13 in Cameron and Trivedi helped me understand how to use bootstrapping, but not when I should consider using bootstrapping.


You use MLE when you cannot apply linear modeling techniques to solve your estimation problem and you are willing to make a distributional assumption about the error process(es) in the model. For instance, most estimation techniques for limited dependent variable models are implemented in Stata and other software as ML estimators. ML estimators are also used in other models that are nonlinear in the parameters. Alternatives would be some form of nonlinear least squares (nl in Stata) or generalized method of moments (gmm in Stata). NLLS and GMM estimators can be implemented with weaker distributional assumptions on the error process(es).

You use bootstrapping when you doubt the validity of the precision estimates (VCE) computed by classical means: for instance, the ordinary, robust, or cluster-robust standard errors produced in any estimation. Bootstrapping is a way of verifying that the classical VCE estimates are sensible, and is often used where there is no analytical solution for the VCE.

Kit

Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                              An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
   An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


*
*   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/