Re: Re: st: Range Merging

[Malcolm Wardlaw <malcolm@mail.utexas.edu>]

I get this response sometimes.   It's simply because this isn't the  
convention for doing long run event studies.  You need -all- the  
companies within that range.  Then you collapse them into an average  
value (a step I left out on purpose).  If you don't pull the entire  
range then you tend to skew the weights on the comparable statistics  
when you deal with different industries.

However this ranged matching issue is a recurring problem for me.  I  
know that in some of the instances I can get around it by doing  
something else, but most of the time, this is the question that I'm  
wrestling with.  I have to match one-to-many in a range.  There's no  
other way around it.

Oh, and I'm aware that there are some syntax problems in my "sample"  
program.  I was rewriting the example to better illustrate my point  
and didn't actually run the program.  Sorry if there was any confusion.

>Malcolm Wardlaw <malcolm@mail.utexas.edu>:
>Reading your post again, it seems clear to me that you should use some
>variety of matching program; what do you gain by matching a company to
>*all* other companies in the universe in some bin, as opposed to
>matching to the nearest k neighbors using e.g. -nnmatch- or
>-psmatch2-?



*
*   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/