appropriate summary statistics for model interpretation/comparison

pinskylab / dynamic_range_model

dynamic range models and forecast methods

2 stars 0 forks source link

Open afredston opened 4 years ago

afredston commented 4 years ago

How do we want to present and compare results from models?

(EDIT: especially need to consider that our model forecasts abundance whereas traditional SDMs predict probability of occurrence)

mpinsky commented 4 years ago

There are two kinds of summary statistics:

Those that are used for model fitting (comparing observations to models)
Those that compare among models. For species distributions, AUC, TSS, and r2 are all useful. Also see Liu, C., M. White, and G. R. Newell. 2011. Measuring and comparing the accuracy of species distribution models with presence absence data. Ecography 34:232–243 and related papers.

afredston commented 3 years ago

leaving this issue open because although we aren't using ABC anymore, we will still need to come up with benchmarks for model comparison.