Am I doing it right for variance component analysis ?

[Meekang Sung]

Hello !

I am new to VCA and this forum, please let me know if I made any mistakes.

I am trying to run a variance component analysis with various levels to see how each level would explain the total variance
There are three measurements for each individual, with a sample size of about 90000


I followed these steps for now
1. Set the data to a long format with measurement
2. set panel structure for individual(caseid), cluster(v001), region(sdist), state(v101)
3. for two-level model for “measurements” at Level-1 nested within “individual” at Level-2, I coded it like
xtmixed measurement || caseid:, var
the purpose of the two-level model is trying to understand the of intra-individual variance.

I am thinking the three-level code to be
xtmixed measurement || v001:|| caseid:,var

I am curious if I should use xtreg instead of xtmixed (or mixed) ?
or if I should reml option instead of var ?

I am doing the analysis all over again because I accidentally messed up with the order of levels
In my short previous experience, convergence was made within an hour or went on for more than 12hrs
I break the calculation in such situations, but perhaps I should wait some more?

Thank you so much for reading this through.

Best,