search

kaitlyngaynor / gorongosa-mesocarnivores

2 stars 0 forks source link

new warnings after workaround #77

Closed klg-2016 closed 4 years ago

klg-2016 commented 4 years ago

https://github.com/kaitlyngaynor/gorongosa-mesocarnivores/blob/2e63e1413ffaad42fed9d990b9655ae98f388978/scripts/02-multi-species-occupancy-analysis-GNP.Rmd#L345

So the only thing I've changed since the last time I successfully ran these models is the workaround suggested by Ken, and now fit13 (which is one of the models I was planning to present) both throws the NaN warning where it hadn't before and has a very different AIC (9920 vs 9895 before). fit12 throws a "model does not converge" warning where it was fine before, and fit11 is fine where it had thrown an NaN warning before, with an AIC different from before by ~10.

The first 3 models I transferred to my table (fit00, fit3, fit001) all ran fine and threw no new warnings.

Any thoughts? If you have a min, would you see if these same warnings come up for you?

I'm trying to pull together the model results to put them all in a table, which is on its way to being a beast.

klg-2016 commented 4 years ago

I could put the detection covariates back into the siteCovs for running these models? and only have them in obsCovs for running the prediction function for making the marginal occupancy figures?

kaitlyngaynor commented 4 years ago

Yeah I might just recommend doing what you suggest in your second comment! just keep em in siteCovs

On Fri, Aug 21, 2020 at 11:46 AM klg-2016 notifications@github.com wrote:

I could put the detection covariates back into the siteCovs for running these models? and only have them in obsCovs for running the prediction function for making the marginal occupancy figures?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kaitlyngaynor/gorongosa-mesocarnivores/issues/77#issuecomment-678436600, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHA7WTZBDR7MS2CMF3OC2TLSB26JZANCNFSM4QHR7EWQ .

klg-2016 commented 4 years ago

I just checked some of the type 3 models as well, and the AIC is off by 4 for fit21, and the model now won't run for fit22 :/ for fit 23, the delta AIC from old to new is 6 and the model now doesn't converge and throws the NaN warning, instead of only the NaN warning.

Cool, will do! I thought I was just going to be transferring results to a table and found a whole bunch of "broken" models, so it threw me off. Thanks for your quick response!

kaitlyngaynor commented 4 years ago

Maybe you just keep it in sitecovs for all model fitting. Then just make a backup model with it in detcovs ONLY for plotting.

On Fri, Aug 21, 2020 at 12:00 klg-2016 notifications@github.com wrote:

I just checked some of the type 3 models as well, and the AIC is off by 4 for fit21, and the model now won't run for fit22 :/ for fit 23, the delta AIC from old to new is 6 and the model now doesn't converge and throws the NaN warning, instead of only the NaN warning.

Cool, will do! I thought I was just going to be transferring results to a table and found a whole bunch of "broken" models, so it threw me off. Thanks for your quick response!

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/kaitlyngaynor/gorongosa-mesocarnivores/issues/77#issuecomment-678442733, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHA7WT5IKEKG2OBRWTYFTO3SB274NANCNFSM4QHR7EWQ .

-- Sent from my phone

kaitlyngaynor commented 4 years ago

I kind of assumed that’s what you were already doing haha. I don’t think you should redo em all! Sorry for not clarifying sooner

On Fri, Aug 21, 2020 at 12:01 Kaitlyn Gaynor gaynor@nceas.ucsb.edu wrote:

Maybe you just keep it in sitecovs for all model fitting. Then just make a backup model with it in detcovs ONLY for plotting.

On Fri, Aug 21, 2020 at 12:00 klg-2016 notifications@github.com wrote:

I just checked some of the type 3 models as well, and the AIC is off by 4 for fit21, and the model now won't run for fit22 :/ for fit 23, the delta AIC from old to new is 6 and the model now doesn't converge and throws the NaN warning, instead of only the NaN warning.

Cool, will do! I thought I was just going to be transferring results to a table and found a whole bunch of "broken" models, so it threw me off. Thanks for your quick response!

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/kaitlyngaynor/gorongosa-mesocarnivores/issues/77#issuecomment-678442733, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHA7WT5IKEKG2OBRWTYFTO3SB274NANCNFSM4QHR7EWQ .

-- Sent from my phone

-- Sent from my phone

klg-2016 commented 4 years ago

Perfect, that's what I'm implementing now. I didn't even think about it/think it would cause any problems. thank you!