Dear all,

I am trying to fix convergence problem of my model using starting values as suggested many times here on statalist. However, I am always getting the following error:

a 0 x 0 is not a vector
an error occurred when mi estimate executed melogit on m=1
r(503)
simplified example of my models and starting values usage:

Code:
mi estimate, dots cmdok: melogit volba X Y Z [pweight=VAHA] || okres_cd:, covariance(unstructured) noestimate

mat a=e(b)

mi estimate, dots cmdok: melogit volba X Y Z [pweight=VAHA] || okres_cd: Z, covariance(unstructured) from(a)
My version of Stata is 16.1 MP.

Could you tell me where the mistake is and give me a suggestion of code change please? Thanks in advance!