Pairwise Correlation between Groups within a Variable

[Vinh Ng]

Dear all,

Suppose I have the following dataset:

clear
input str7 yq   int firmid float  retn
2000q1     1    0.2564
2000q2     1    0.2700
2000q3     1    0.2160
2000q4     1    0.1005
2000q1     2    0.2493
2000q2     2    0.3820
2000q3     2    0.2179
2000q4     2    0.3778
2000q1     3    0.9692
2000q2     3    0.4204
2000q3     3    0.9001
2000q4     3    0.2245
2000q1     4    0.7196
2000q2     4    0.3100
2000q3     4    0.5980
2000q4     4    0.4791
2000q1     5    0.1168
2000q2     5    0.2005
2000q3     5    0.1802
2000q4     5    0.2783
end

I'd like to calculate the pairwise correlation between retn for each pair of firms (e.g., correlation between firmid 1's retn and firmid 2's retn ) and store correlation coefficients in separate variables.

Could anyone please help me with this?

Thank you very much for your help,

Vinh

[Clyde Schechter]

Each pairwise correlation is a single number, so it makes little sense to waste a whole variable on each. The natural way to represent these pariwise correlations would be as a matrix.

To get there, this is one of the few situations where a wide data layout works better.

reshape wide retn, i(yq) j(firmid)
pwcorr retn*
matrix wanted = r(C)

[Vinh Ng]

Thank you so much for your help, Clyde. Your code works perfectly.

Vinh