r - Test for Heteroscedasticity on a caret nnet model -
is there function within caret (or package) can perform breusch-pagan / cook-weisberg test heteroskedasticity on 'nnet' model trained using caret?
e.g. similar library(car); ncvtest or library(lmtest); bptest lm objects, works on nnet objects created caret?
example data
library(caret) set.seed(4) n <- 100 x1i <- rnorm(n) x2i <- rnorm(n) yi <- rnorm(n) dat <- data.frame(yi, x1i, x2i) mod <- train(yi ~., data=dat, method="nnet", trace=false, linout=true) this produces plot of fitted vs residuals:

no there not in package right now.
Comments
Post a Comment