st: dyads - correncting for multiple source of autocorrelations

[fabio.zona@unibocconi.it]

Dear statalist

I have a problem with dyadic data, i.e. relations between pairs of firms.
Due to dyadic structure of data, I have a problem with autocorrelation.

Quadratic Assignment Procedure – a simulation based approach – allows to simulate and estimate correct standard errors. However, this test is based on simulation and does not calculate any variance-covariance matrix of estimates.

I do NEED this variance-covariance matrix of estimates. Hence, I have to correct for autocorrelation.
Clustering might be one option. Clustering allows correction of correlation within each group, reported in the variable “group” used for clustering, i.e., cluster (group).

However, with dyadic data I have multiple sources of correlation. Below I represent  - in a square matrix and in a triangular structure – all possible dyads of 5 firms, A, B, C, D and E. 

I might set the “group” for clustering around the vertical dyads; for example, I assign the number 2 to all dyads with firm B (column which starts with B): then stata corrects for autocorrelation within this group numbered 2 (dyada BA, BB,BC,BD,BE). 

However, stata would treat the dyads BE  and AE  as independent, because – from the vertical perspective – BE and AE belongs to different groups. In fact, BE and AE are not independent, because they share E as a common source of correlation.
In other words, I have multiple source of correlation to be adjusted, the row and the column simultaneously.

How is it possible to resolve this problem?


   A   B   C   D   E
A      1   0   1   1 
B  1       1   1   1
C  0   1       0   0
D  0   1   0       1 
E  1   1   0   1  



   A   B   C   D   E
A       
B  1       
C  0   1    
D  0   1   0        
E  1   1   0   1  


 

*
*   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/