Interpreting "estat vce, corr" output -- post-estimation of menbreg with random slope and intercept

[Abby Em]

Hi everyone,
I am interested in getting some assistance interpreting the output of a estat vce, corr command used for post-estimation of a mixed model with a random slope and a random intercept. Specifically, I am interested in knowing which number represents the correlation coefficient for the covariance between the random intercept and random slope. I have not been able to find any annotated output for this command with this specific model type. I am happy to provide additional details on the data and code - whatever would be helpful in answering my question.
Thanks very much!

menbreg DrugNumber c.admdis || mrn: admdis, vce(robust) irr covariance(unstructured)
estat vce, corr

Here are mock-ups of what the output looks like on Stata, for the model:

Here is what the output looks like for the estat vce, corr command used following the above model:

[Clyde Schechter]

What you're looking for doesn't show anywhere in the outputs you've got. The closest thing to it is the covariance between the random intercepts and the random slopes. That's in the leftmost column of the bottom row of the -mixed- output (not the -estat vce- output). But you want the correlation. So take that covariance and divide it by the product of the square roots of the "coefficients" of var(admdis) and var(_cons) from the -menbreg- output.

[Abby Em]

Hi Clyde, Thank you so much for your response. When I follow your suggested procedure, I get an answer of -1. (I've also tried generating BLUPs for the random slope and random intercept using predict stub*, reffects and then did did a corr command between these two variables and also got -1.0). Would you have any ideas, in terms of model set-up (or conceptually) why I would be getting -1.0? Thanks again!

[Clyde Schechter]

That's not a question I can answer from afar. The correlation between intercepts and slopes is a complicated thing that depends not only on the model overall, but also on how the variables involved are centered. I think you would be better off seeking assistance from somebody at your workplace (or school if you are a student) who can sit down with you, having access to the complete data set, and work through how you got to where you are and figure this out.