search

SCBI-ForestGEO / McGregor_climate-sensitivity-variation

repository for linking the climate sensitity of tree growth (derived from cores) to functional traits
0 stars 0 forks source link

try loosening criteria for candidate variables #55

Closed teixeirak closed 5 years ago

teixeirak commented 5 years ago

@mcgregorian1, could you please try loosing the criteria for candidate variables to a threshold of dAIC =1, or positive dAIC in at least 2 models? (This gives the same set.) I'm concerned that the criteria is a bit too strict. TLP improves (but not significantly) 3 of the 4 single-variable models yet is excluded from consideration in the full models based on current criteria. This contrasts with LMA and WD, which make 3-4 models worse.

mcgregorian1 commented 5 years ago

I can do that. If we do dAIC >= 0 then that would include LMA and WD, whereas if we keep the threshold at dAIC >=1 we would only capture the addition of TLP, which is what we want.

I've run the code again and updated the files

teixeirak commented 5 years ago

Thanks! Interesting that TLP comes out in all of the very top models except 1999, and at least one top model for each year -- confirmation that including it is the right approach. It looks like you forgot to include it in tested_traits_best_coef.csv. (Not sure that table updated correctly... also seems to be missing 1999.)

mcgregorian1 commented 5 years ago

whoops my bad, there was a part of my code that wasn't keeping all the variables in, plus I hadn't realized the full list of best models had increased that much. It's fixed now

teixeirak commented 5 years ago

Thanks! The analyses all look good; let's count these as "final" and proceed with preparing the manuscript.