hi,

I am a R-novice. I want to do some outlier cleaning and over-all-scaling from 0 to 1 before putting the sample into a random forest.

g<-c(1000,60,50,60,50,40,50,60,70,60,40,70,50,60,50,70,10)

If i do a simple scaling from 0 - 1 the result would be:

> round((g - min(g))/abs(max(g) - min(g)),1)

 [1] 1.0 0.1 0.0 0.1 0.0 0.0 0.0 0.1 0.1 0.1 0.0 0.1 0.0 0.1 0.0 0.1 0.0

So my idea is to replace the values of each column that are greater than the 0.95-quantile with the next value smaller than the 0.95-quantile - and the same for the 0.05-quantile.

So the pre-scaled result would be:

g<-c(**70**,60,50,60,50,40,50,60,70,60,40,70,50,60,50,70,**40**)

and scaled:

> round((g - min(g))/abs(max(g) - min(g)),1)

 [1] 1.0 0.7 0.3 0.7 0.3 0.0 0.3 0.7 1.0 0.7 0.0 1.0 0.3 0.7 0.3 1.0 0.0

I need this formula for a whole dataframe, so the functional implementation within R should be something like:

> apply(c, 2, function(x) x[x`<quantile(x, 0.95)]`<-max(x[x, ... max without the quantile(x, 0.95))

Can anyone help?

Spoken beside: if there exists a function that does this job directly, please let me know. I already checked out cut and cut2. cut fails because of not-unique breaks; cut2 would work, but only gives back string values or the mean value, and I need a numeric vector from 0 - 1.

for trial:

a<-c(100,6,5,6,5,4,5,6,7,6,4,7,5,6,5,7,1)

b<-c(1000,60,50,60,50,40,50,60,70,60,40,70,50,60,50,70,10)

c<-cbind(a,b)

c<-as.data.frame(c)

Regards and thanks for help,

Rainer