Generating Pairwise Covariances from a single vector across different individuals

[Chinmay Sharma]

Hi All,

I know the title of the question seems confusing, but I will try to explain my problem to the best of my ability.
I currently have a dataset that looks like the following:

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(Identifier Year Score)
1 1990    2
1 1991   43
1 1992    3
1 1993    4
2 1990   23
2 1991    4
2 1992   32
2 1993   43
3 1990 3234
3 1991   23
3 1992   32
3 1993   32
.    .    .
.    .    .
end

------------------ copy up to and including the previous line ------------------


In the dataset above, we have three vectors: the first one corresponds to an individual's identifier, the second corresponds to the year of the observation, and the third corresponds to the score that an individaul obtained. What I wish to do is to basically calculate a vector of covariances of scores for each possible pair of individuals (this will yield nC2 different unique combinations) across different years. Basically, I want to obtain the time-covariance of each possible (i,j) combination, with each term in the covariance calculation corresponding to the particular (demeaned) product of score that i and j obtained in that year, and then dividing by the number of years that (i,j) both have scores.

Any help on this is much appreciated.

Best Wishes,
Chinmay

[Jesse Wursten]

I think you might be able to do this using -pwcorrf Score, reshape-
Alternatively, you can reshape wide the data yourself and use the official pwcorr command.