search

ck37 / varimpact

Variable importance through targeted causal inference, with Alan Hubbard
57 stars 13 forks source link

Error #4

Closed ahubb40 closed 7 years ago

ahubb40 commented 7 years ago

I get the following error.

numValues: 12, numResults: 12, stopped: TRUE calling combine function evaluating call object to combine results: fun(accum, result.1, result.2, result.3, result.4, result.5, result.6, result.7, result.8, result.9, result.10, result.12) returning status TRUE Error in { : task 11 failed - "attempt to select less than one element in get1index"

ahubb40 commented 7 years ago

Follow-up, all variables get "returning status FALSE", so perhaps it is an error if there are no variables that get through screening process.

ck37 commented 7 years ago

Hmm sorry about that, what does your SuperLearner library look like?

ahubb40 commented 7 years ago

SL.library <- c("SL.glmnet","SL.randomForest","SL.mean")

ahubb40 commented 7 years ago

But, I'm guessing it never gets to that because of the screening. It's very small sample sizes, so wouldn't surprise me.

ck37 commented 7 years ago

It looks like the 11th variable in that section (either factors or numerics) is what is causing the error - if you run with verbose = T and verbose_tmle = T does it provide more details around that variable?

ahubb40 commented 7 years ago

I get same output when I do that. I'm not sure it ever gets to the TMLE part.

ck37 commented 7 years ago

Yep I hear you on that. I think I will need to dig into the code as it's running to figure this out - is it possible to send over the data & script?

Also do you know if you're running this in parallel or not? Because it will usually hide a lot of the debugging output when run in parallel. If you specify parallel = F that will ensure that it runs sequentially.

ck37 commented 7 years ago

Closing this issue since the parallel backend is now future; I believe that the issue is likely fixed.