mixed command: Random Slope with cov(independent) structure

[Vincent KH Lee]

Hi,

I'm concerned whether random slope model with cov(independent) makes sense or not.

1.
mixed y x1##x2, || Country: x2, cov(independent) 

2. 
mixed y x1##x2, || Country: x2, cov(unstructured)

From what I've understood, for both #1 and #2, whether I set the covariance structure to independent or unstructured, they are both random slope.

However, I've never seen a random slope example with cov(independent).

For #1, Stata returns the result within 4 iterations, whereas with #2 the model never converges, with message "not concave", and "backed up"

cov(unstructured) works if x2 is not included as Country-level covariate, which doesn't really make sense theoretically in my case.


So my question is:

Is it still a random slope model (#1) with cov(independent) option?

And is it okay to interpret and write out multi-level equations based on #1 or should I consider it as random intercept model?

Thank you.