我在估算cross-covariance matrix時遇到了困難，希望能得到幫助。
假設已有dataset：

company	date	stock return
a1	DATE#1	RET(a1,1)
a1	DATE#2	RET(a1,2)
...	...	...
a1	DATE#T	RET(a1,T)
a2	DATE#1	RET(a2,1)
...	....	....
a2	DATE#T	RET(a2,T)
...	...	...
a5000	DATE#1	RET(a5000,1)
...	...	...
a5000	DATE#T	RET(a5000,T)

請問該如何利用PROC TIMESERIES，估算這5000家公司股票報酬率的cross-covariance matrix呢?

cross-covariance:
GAMMA(K)=E[(Xi,t - E(Xi))(Xj,t+K - E(Xj))],
i=1,2,...,N;
j=1,2,...,N;
N: total number of companies
K: lag period

-----
我嘗試透過PROC IML建立新的dataset(如下)，但遇到memory相關問題

company i	company j	date	return i	return j
...	...	...	...	...