Comparing Models With and Without Spatio-Temporal Random Effects #236
-
|
Is there a way to quantitatively compare the amount of variability explained by a model incorporating spatio-temporal random effects + fixed effects versus a model containing only fixed effects? I know I can plot predictions and conduct cross-validation, but is there any sort of in-sample metric I can derive to supplement these approaches? As an example, I fit spatial and non-spatial models in mgcv recently, and I could compare the deviance explained by the two versions. Thanks! |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments
-
|
@Nathan-Hebert, sorry I completely missed this question back in July! This work on an R2 function might be what you're looking for: https://github.com/pbs-assess/sdmTMB/blob/main/scratch/r2-glmm.qmd within this issue. Not all families are in there yet and it's not thoroughly tested, but that might be a start. |
Beta Was this translation helpful? Give feedback.
-
|
Awesome, thank you! |
Beta Was this translation helpful? Give feedback.
@Nathan-Hebert, sorry I completely missed this question back in July! This work on an R2 function might be what you're looking for: https://github.com/pbs-assess/sdmTMB/blob/main/scratch/r2-glmm.qmd within this issue.
Not all families are in there yet and it's not thoroughly tested, but that might be a start.