Optimize box-cox conversion, function cannot be evaluated by initial parameters

Question

I am trying to do an automatic box-cox conversion (which should be generally useful for normal data for people), but with a mistake formulate my optimization in such a way that R optim is fine. This usually works, but I don’t understand why it fails on variables with extreme tilt.

The idea is to select the Lambda parameter in the box-cox transform, which minimizes the absolute amount of data set distortion.

library(car)
library(moments)

xskew <- function(data,par){
    abs(skewness(bcPower(data,lambda=par[1]),na.rm=T)) # minimize abs(skew)
}

boxit <- function(x){
    res <- optim(par=c(-5,5), xskew, data=x+1)         # find argmin(^) lambda
    print(res$par)
    return(bcPower(x+1,lambda=res$par[1]))

This usually works well, for example:

> skewness(rbeta(1000,12,3))
[1] -0.6439532

becomes

> skewness(boxit(rbeta(1000,12,3)))
[1] -5.980757e-08

--- almost 0 skew.

But on one extremely skewed variable, I get:

Error in optim(par = c(-5, 5), xskew, data = x + 1) (from #2) : 
  function cannot be evaluated at initial parameters

My thoughts are probably the following:

• Do not catch how bcPower works with values ​​near zero or infinity.
• , - , , .

Answer 1

, 2 1 . , :

optim(par=-2, xskew, x=x, method="Brent", lower=-20, upper=20)

xskew.