Variance Covariance Matrix for Panel Data for each Year

[Vince Williams]

Dear Stata community:
I have the following simple question. Suppose I have panel data and 10 variables. My data is from year 1989 to 2016.
How can I compute the variance-covariance matrix for each year and store it as a variable?
I know if I use:
correlate d1-d10, covariance
I get variance-covariance for all the years from 1989-2016.

CompanyID	YEAR	d1	d2	d3	d4	d5	d6	d7	d8	d9	d10
1	1989	0.1	3	3	3	3	20	21	22	23	24
1	1990	5	2	2	2	5	20	19	18	17	16
1	1991	6	2	2	2	6	20	25	30	35	40
2	1989	20	2	2	2	7	20	19	18	17	16
2	1990	7	2	2	2	8	20	19	18	17	16
2	1991	10	2	2	2	9	20	19	18	17	16

[Vince Williams]

I do:
search corrmat
Is this package going to work?

[Phil Bromiley]

I have no idea what your second post means. You certainly can set up a loop over years, calculate the coverance matrix for each year and then do whatever you want with that matrix.

That is something like:
forvalues t=2001/2006 {
corr d* if year==`t'
**need to write matrix somewhere here**
}

Alternatively, it looks like you could loop and use corrmat instead of corr.

I don't know how you want to write the covariance matrix. You could save the matrices as matrices (simply given the a base name and put the `t' at the end of it: matrix corr`t'=r(C)) or use corrmat's facility to save a matrix. But if you want to store it as a variable you'll have to figure how you're going to do that. If you want to write these in different variables (i.e. put each element in a different variable but stack the years one after the other, you can generate the needed variables before the loop as missing (i.e., g corrd1d2=.) and then replace the value for each year.