Re: st: RE: RE: RE: Imputing values for categorical data

[Richard Williams <Richard.A.Williams.5@nd.edu>]

At 10:08 AM 4/16/2004 -0400, Richard Goldstein wrote:
> There is a problem with the indicator variable method for missing 
> values.  See Jones, MP, (1996) "Indicator and Stratification Methods for 
> Missing Explanatory variables in Multiple Linear Regression," JASA, 91: 
> 222-230.
>
> Rich Goldstein
Also, in his Sage monograph, "Missing Data" (a great book I might add), 
Paul Allison presents a pretty devastating critique of the MD indicator 
method, saying that, while it is "remarkably simple and intuitively 
appealing", it unfortunately "generally produces biased estimates of the 
coefficients".  He shows that listwise deletion is superior to this method.

Cohen and Cohen advocated the method in the 1975 edition of their book, 
"Applied multiple regression/correlation analysis for the behavioral 
sciences."  In the 2003 edition, they pretty much abandoned it.

I taught and advocated this approach for years.  I let my students have 
access to old exams, and I have to tell them that, every time I said to use 
this method, I was wrong!


*
*   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/