search

pfmc-assessments / VASTWestCoast

VAST for the NWFSC West Coast data
Other
5 stars 1 forks source link

Model selection #27

Open kellijohnson-NOAA opened 3 years ago

kellijohnson-NOAA commented 3 years ago

Write a protocol for model selection for a given species when looking into different distributions or other configurations of VAST. Currently, we are using AIC but cross-validation methods could also be used.

James-Thorson-NOAA commented 3 years ago

I don't have a ton of trust in marginal AIC; I sent Kelli the paper accepted at Fisheries Research (and will add the link here once its available online) showing that it often effects the delta-lognormal even in cases when it has a poor agreement about index scale relative to a design-based estimator.