search

jrgant / FluHospPrediction

R package repository to reproduce the flu hospitalization prediction project results
1 stars 0 forks source link

GAM Errors #10

Closed jrgant closed 4 years ago

jrgant commented 4 years ago

The GAM models are throwing errors regarding incorrect degrees of freedom, resulting in their being dropped from the learning stack. May need to set function arguments explicitly for each one.

jrgant commented 4 years ago

I'm not sure why, but I haven't seen this error message again. Checked test runs, and the GAM learners returned as trained.

Make sure to check results after the OSCAR jobs run.

(I dropped the learner_fits objects from the saved output to reduce filesize. Consider pulling the trained statuses from here if they're not currently being written to results.)

jrgant commented 4 years ago

Error message is still there. Written to job logs on slurm.

jrgant commented 4 years ago

Issue most likely is that template (categorical fold ID) was accidentally being included as a covariate. Made a fix in fhp_make_task() but need to test it.

jrgant commented 4 years ago

Removed template, template_numeric, and cid. No dice.

jrgant commented 4 years ago

Probably too many parameters in the model. Maybe should incorporate each GAM into a screening algorithm or check how to penalize too many model parameters.

jrgant commented 4 years ago

Fixed