1

I can not find the p values for multinominal logit model using caret package

Calculating p values are provided here but multinominal distribution should be binominal? does any one knows the right distribution and p value calculation for multi nominal? (pvalues significance are not the same if i follow ucla.edu in r OR if i do multinominal in stata) many thanks in advance

assuming the following, reproducible example, how to get p values from caret in summary function 

data("Fishing", package = "mlogit") ;head(Fishing)
set.seed(87)
inTrainingSet <- createDataPartition(Fishing$mode,p=.80,list=0)
Train <- Fishing[inTrainingSet,]
Test  <- Fishing[-inTrainingSet,]
myControl <- trainControl(method = "cv", 
                          number=10,
                          verboseIter = TRUE)


yy <- factor(Train$mode,levels = c( "beach", "pier", "boat","charter"))
xx <- Train[,2:10]

mmulti <- train(x=xx,
                 y=yy,
                 method = "multinom",
                 trControl = myControl,
                 tuneLength = 10); summary(mmulti) 

multi.rrr = exp(t(coef(summary(mmulti)))) ; multi.rrr
NelsonGon
  • 13,015
  • 7
  • 27
  • 57
Seyma Kalay
  • 2,037
  • 10
  • 22
  • caret is not the right package to be used for this. Also `method = 'multinom'` uses a neural network whic his probably not something you wish to use. [Source](https://rdrr.io/rforge/caret/man/models.html) – missuse Aug 23 '19 at 19:12

0 Answers0