CODEWITHSUNDEEP
rrandomkernel-density
jpcgandre
jpcgandre

Reputation: 1505

Different results when generating random samples from kernel density

library(ks)

x<-rnorm(1000)
hist(x, col="red")

y <- rkde(kde(x), n=1000)
hist(y, col="green")

y <- rkde(density(x), n=1000)
hist(y, col="blue")

The last histogram is way wrong. I've used density before and I've found that it was accurate for far more complicated distributions. Why in this case it performs so badly? Thanks

Upvotes: 0

Views: 261

Answers (1)

Dason
Dason

Reputation: 62003

Because you're using the function wrong. rkde expects an object of the class kde. density doesn't return a kde object and is structured differently.

It would be like telling somebody to shoot their pistol and handing them shotgun shells and then wondering why when they fired it didn't really do anything.

Upvotes: 3

Related Questions