Maxent do not works with factors

[FVFaleiro]

Hi Babak,

I make some tests and when maxent use categories of factors with letters it do not works, but if I use numbers as factors it works fine.
Check it bellow.

Cheers!

library(sdm)

data

train1 <- data.frame(sp = rep(0:1, each = 50), n1 = rnorm(100), n2 = rnorm(100), f1 = factor(rep(c("a", "b"), 50)))
train2 <- data.frame(sp = rep(0:1, each = 50), n1 = rnorm(100), n2 = rnorm(100), f1 = factor(rep(c(1, 2), 50)))
train3 <- data.frame(sp = rep(0:1, each = 50), n1 = rnorm(100), n2 = rnorm(100))

format

d1 <- sdmData(formula = ~., train = train1)
d2 <- sdmData(formula = ~., train = train2)
d3 <- sdmData(formula = ~., train = train3)

fit and test models

m1 <- sdm(~., data = d1, 
          methods = 'maxent',
          replication = 'cross-validation', cv.folds = 10)
m2 <- sdm(~., data = d2, 
          methods = 'maxent',
          replication = 'cross-validation', cv.folds = 10)
m3 <- sdm(~., data = d3, 
          methods = 'maxent',
          replication = 'cross-validation', cv.folds = 10)

check models

m1 # factors as letters
m2 # factors as numbers
m3 # only numeric predictors

[babaknaimi]

Hi, Thanks for reporting this.
My initial guess is that the factor variables should be defined as an argument (model setting) to the algorithm that is handled by the maxent function in the dismo package. I think if you define which variables are categorical as the model argument [by adding modelSettings=list(maxent=list(factors=c("f1"))) to the sdm function] it would fix the issue.

I will check to make this automatically handled.