Have completed a successful run of Stata expectation maximization estimation of the covariances of 94 variables. I used _getcovcorr to further examine the matrix, and to check on whether the matrix is positive semi-definite and/or positive definite. The name of my matrix is Cov_em. To test, I issued these commands according to the documentation for _getcovcorr:
_getcovcorr Cov_em, check(psd)
_getcovcorr Cov_em, check(pd)
The result of these commands is simply a dot (period), without any statement or message regarding whether the matrix passed the tests. Thus, the result is ambiguous. Does "nothing" returned except a period mean these tests were passed?
I then issued the command:
matrix list r(C) in case a message was embedded therein. The matrix was correctly listed, but no information was described regarding whether the tests were passed.
I next issued the commands:
matrix symeigen X v = Cov_em
matrix list v
I got 94 eigenvalues printed out; the last, smallest eigenvalue was: .00026395 a positive number
The matrix determinant was a very small number, 9.580 e -95, indeed a very small positive number.
Fellow statisticians, what do we have here? A positive definite matrix? A positive semi-definite matrix?
Pete sends his best regards to all with thanks for your opinions.
_getcovcorr Cov_em, check(psd)
_getcovcorr Cov_em, check(pd)
The result of these commands is simply a dot (period), without any statement or message regarding whether the matrix passed the tests. Thus, the result is ambiguous. Does "nothing" returned except a period mean these tests were passed?
I then issued the command:
matrix list r(C) in case a message was embedded therein. The matrix was correctly listed, but no information was described regarding whether the tests were passed.
I next issued the commands:
matrix symeigen X v = Cov_em
matrix list v
I got 94 eigenvalues printed out; the last, smallest eigenvalue was: .00026395 a positive number
The matrix determinant was a very small number, 9.580 e -95, indeed a very small positive number.
Fellow statisticians, what do we have here? A positive definite matrix? A positive semi-definite matrix?
Pete sends his best regards to all with thanks for your opinions.