[R] Winsorized mean and variance

[Roberto Perdisci]

This is of great help, thanks!

Roberto

On Thu, Aug 27, 2009 at 7:20 AM, Jim Lemon<jim at bitwrit.com.au> wrote:
> Roberto Perdisci wrote:
>>
>> Hello everybody,
>>  after searching around for quite some time, I haven't been able to
>> find a package that provides a function to compute the Windorized mean
>> and variance. Also I haven't found a function that computes the
>> trimmed variance. Is there any such package around?
>>
>>
>
> Hi Roberto,
> The Winsorized variance is similar to the trimmed variance, except that the
> extreme values are substituted rather than dropped. Define the quantiles
> within which you want to retain the original values and then substitute the
> values at the quantiles for all values more extreme in the respective sign
> direction. Like this:
>
> testdat<-rnorm(20)
> winsorVar<-function(x,probs=c(0.05,0.95)) {
> xq<-quantile(x,probs=probs)
> x[x < xq[1]]<-xq[1]
> x[x > xq[2]]<-xq[2]
> return(var(x))
> }
>
> Jim
>
>