Dear everyone,
I'm running a three-level regression using meglm, and I'm getting extremely high residual variances (~10E+08) at level 2 and 3. I'm trying to understand why I'm getting such high variance and how to fix this.
For my model, I am using the following code:
meglm h11 [pw=v005] || v000: || v001:, link(logit) family(bernoulli)
where h11 is a binary (0 or 1) outcome variable, v000 and v001 are variables indicating the levels, and v005 is the weight variable at level 1. My sample size is ~300,000 at level 1, ~2000 at level 2, and 30 at level 3. I'm getting similar residual variance when I include 5+ other explanatory variables.
Any comments would be much appreciated. Thank you.
I'm running a three-level regression using meglm, and I'm getting extremely high residual variances (~10E+08) at level 2 and 3. I'm trying to understand why I'm getting such high variance and how to fix this.
For my model, I am using the following code:
meglm h11 [pw=v005] || v000: || v001:, link(logit) family(bernoulli)
where h11 is a binary (0 or 1) outcome variable, v000 and v001 are variables indicating the levels, and v005 is the weight variable at level 1. My sample size is ~300,000 at level 1, ~2000 at level 2, and 30 at level 3. I'm getting similar residual variance when I include 5+ other explanatory variables.
Any comments would be much appreciated. Thank you.