I'm comparing two treatments using survival time. The model is lognormal and has an intercept beta1, coefficient beta2 on size and beta3 on treatment.
I have issues with convergence of mcmc chains on beta1 and beta3.
My question: is there convergence on chains of beta1 and beta3?
Yeah, I get that. But if we can only see 2% of estimates can a ruling be made. In my limited experience, my chains converge pretty quickly, but my models are fairly simple. We also don't know if thinning will still be applied to these. And how do models come up with effective sample numbers, is that applicable here?