[ADMB Users] Getting the likelihood from a Random Effects model
hskaug at gmail.com
Thu Oct 29 02:13:14 PDT 2009
No, I do not think there is something like the hat matrix.
In fact, I am not sure what the hat-matrix amounts
to outside linear models, so I cannot say if
it can easily be calculated from what ADMB-RE returns.
My view is that most type of diagnostics and model selection
criteria should be calculated post-hoc, outside ADMB, for instance
in R. However, the user manual should explain (better than
it does currrently) how to find the needed quantities in
the ADMB output.
> Presumably AIC criterion also depends somewhat on the "level of
> focus", i.e. whether you are interested in population- or
> individual-level prediction (e.g. see Vaida and Blanchard 2005): does
> AMDB-RE quantify "trace of the hat matrix" or equivalent?
More information about the Users