Run the models without pooling ...

lizzieinvancouver / ospree

Budbreak review paper database

3 stars 0 forks source link

Run the models without pooling ... #345

Closed lizzieinvancouver closed 2 years ago

lizzieinvancouver commented 4 years ago

@MoralesCastilla Can I ask you to run two more models and see how they compare in R2 and LOO? Could you run (1) eqns 1 and 2 (from your outline) but WITH phylogeny and WITHOUT spp on the intercept (aka eqns 3-4 but drop spp on the intercept)? What I think is an issue here is that both partial pooling and phylogeny correct for the non-independence of species, and I think we should show that without partial pooling you do get an improvement from phylogeny.

MoralesCastilla commented 4 years ago

I have only checked the model coefficients so far and they seem. Still need to check phylosignals and accuracies (R2s and LOOs)

The estimates of the effects of each cue, remain qualitatively similar for eq 1, but do not converge for eq 2

Model eq 1 - Forcing (-6.65) ; Chilling (-5.33) ; Photo (-0.95) - none overlaps with zero
Model eq 2 - it does not converge // crashes

lizzieinvancouver commented 2 years ago

We switched to a PMM model and no longer need this (I believe). @MoralesCastilla re-open if you disagree.