Best way to do model selection and diagnostics (Bayesian and frequentist)

chicas-spatstat-reading-group / chicas-spatstat-reading-group.github.io

Webpage for the CHICAS Spatial Statistics Reading Group

https://chicas-spatstat-reading-group.github.io/

Other

1 stars 0 forks source link

Best way to do model selection and diagnostics (Bayesian and frequentist) #6

Open poppymiller opened 7 years ago

poppymiller commented 7 years ago

Topics could include:

Bayesian model selection
Bayesian model averaging
Stepwise regression vs likelihood ratio tests
Diagnostics for glms and other models where the residuals are expected to look a bit strange, so linear model diagnostics don't really work

drkamarul commented 7 years ago

perhaps also on GAM with spatial data?

RhiannonEdge commented 7 years ago

I often think it is odd that in some cases models may change dependent on the method of parameter selection. I think this stresses the importance of collaboration with subject experts. However, I have struggled to justify this to clinicians who are seeking data-lead conclusions. Perhaps some discussion around this?