ML regression and listwise deletion are different problems. E.g., on SPSS you can have OLS regression with pairwise deletion. It is Stata that, unfortunately (from my perspective), uses listwise deletion. However, to contradict my opinion, I also quote from a recent post on this list "there is a good reason why it [pairwise deletion] hasn't been implemented [in Stata] (and why I don't feel like implementing it): it isn't a good way of dealing with missing data, it is actually worse than listwise deletion however unintuitive that may seem (Allison 2002).
Allison, Paul D. (2002) "Missing Data", Thousand Oaks: Sage."
Having said that, -regmsng- (from STB-47) fits regression with missing observations for some variables. An example in the help of -reghf- (linear regression with multiplicative heteroscedasticity; from STB-47), declares that the command is using "full-information mle".
If none of the is satistactory and you want to program some FIML by yourself, you can have a look at the ado file for -espoisson- (from SJ4-1), -oheckman- (from http://siteresources.worldbank.org/INTPOVRES/Resources), -ivglog- (from ssc), which all declare using FIML.
Hope this somewhat helps,
Nicola
At 02.33 02/06/2007 -0400, you wrote:
>Is it possible to estimate a regression model using ML
>(instead of OLS)? Other programs (e.g., MPLUS, AMOS) allow
>one to estimate regression models using ML (or they refer to
>it as FIML), which is useful when one has missing data on
>observations and one does not want to have one's sample size
>reduced (by listwise). If this is possible, what is the
>synatx one would use?
>
>Thanks,
>John.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
