st: regression with dyadic data, QAP or fixed effects?

[Derek Darves <darves@uoregon.edu>]

Hi All,

I have a dataset that consists of all possible pairs of a group of  
~500 companies (making ~118k dyads).

The variables in the dyadic dataset deal with common occurrences,  
e.g. a variable is set to one if both members of the dyad are members  
of the same organization. Other variables like sales that are  
continuous, and are equal to the geometric mean of the sales for each  
company in the dyad.

This is my question: obviously the SEs of of any regression with this  
data will be downward biased, since many of the errors should be  
correlated (e.g. between to dyads with a common subunit). Does anyone  
have any ideas about how best to adjust the SEs? I have not found an  
implementation of QAP within Stata, so I am wondering if there is  
some way to analyze this data within Stata, e.g. by using a dummy for  
each of the 500 companies that makes up a dyad, or using robust SE  
correction, etc.

The DV is binary, and is set to 1 if both companies share a certain  
attribute.
I have also computed the DV as a count of the number of common  
attributes b/w the two firms in a dyad. Thus, I am interested in  
either logit or nbreg estimation techniques. Umlike the IR literature  
which uses dyads of countries, I do not have multiple years of data,  
just a single cross-section of company dyads.

Any ideas on how to analyze this data would be greatly appreciated.

Derek
*
*   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/