search

SimonDedman / gbm.auto

Machine-learning Boosted Regression Tree software suite for species distribution modelling in R
https://doi.org/10.1371/journal.pone.0188955
Other
18 stars 6 forks source link

Zero Inflation with Non-Default data families. #74

Closed nffarabaugh closed 2 years ago

nffarabaugh commented 2 years ago

Hello I am having a hard time with automated zero-inflation checks with Non-Default data distribution families.

I am working with ecological count data that is better described by Poisson distribution than Gaussian. Many of these data sets are zero inflated, however if I run the families as : fam1 = c("bernoulli", "binomial", "poisson", "laplace", "gaussian"), fam2 = c( "poisson")

I am met with the error : gbm.fit(x = x, y = y, offset = offset, distribution = distribution, : Poisson requires the response to be a positive integer.

If I specify the families as:

fam1 = c(""binomial" ), fam2 = c( "poisson")

It seems like the binomial models are not run at all.

Is there a transformation happening during this process that makes poisson distribution inappropriate for the zero inflation check= TRUE, process?

Thanks for the help

SimonDedman commented 2 years ago

gbm.auto L419 poisson added as an exception.

nffarabaugh commented 2 years ago

Amazing Thanks!