Initial log-likelihood is incorrect when starting variables are not zero

RSGInc / RSGHB

Functions for Hierarchical Bayesian Estimation: A Flexible Approach

Other

25 stars 16 forks source link

Initial log-likelihood is incorrect when starting variables are not zero #80

Open jeffkeller87 opened 8 years ago

jeffkeller87 commented 8 years ago

As per conversation with Stephane Hess.

This only affects the initial diagnostics report—estimation is unaffected. I suspect model$ll0 is correct, but not when using non-normal distributions.

jeffkeller87 commented 8 years ago

Below are the scripts, data, and conversations used to identify the bug.

Hess's Materials.zip