PROC NLP

[deleted_user]

I am optimizing the Bernoulli log-likelihood function using PROC NLP. After running, I do not receive errors, but I do encounter this note:


The specification of SIGSQ= is not valid for this type of covariance
matrix and is used only for the default of the VARDEF= option.


I have double and triple checked my code. SAS documentation states that with MAX=, I need to specify SIGSQ=1 and can use any COV= option. Any ideas?

[Doc_Duke]

Call tech support. If you've read the documentation as completely as you say, then either you've uncovered a bug in NLP or a documentation defect that needs correction. If either of those are true, then tech support is the way to get it addressed.

You may want to check other parts of your code first. I noticed that MAX is a statement, not an option.