Assignment to data.frame causes an unwanted type change in R

Say I generate some data like so:

dat <- data.frame(x = rnorm(100), y = rnorm(100), z = rnorm(100))
dat[sample(nrow(dat), 5), 3] <- NaN
dat[sample(nrow(dat), 5), 3] <- Inf

Now, some of the z-values are Inf or NaN.

The first 10 rows of the results from

cut(dat$z[is.finite(dat$z)],6)[1:10]

are

 [1] (0.286,1.17]   (0.286,1.17]   (0.286,1.17]   (0.286,1.17]   (0.286,1.17]  
 [6] (0.286,1.17]   (-1.48,-0.599] (-1.48,-0.599] (-0.599,0.286] (0.286,1.17]  
6 Levels: (-2.37,-1.48] (-1.48,-0.599] (-0.599,0.286] ... (2.06,2.94]

but if I try to make the following assignment

dat$col[is.finite(dat$z)] <- cut(dat$z[is.finite(dat$z)],6)

I get integers instead of the labels:

> dat$col[1:10]
 [1]  4  4  4  4  4  4  2  2 NA  3

How do I assign the factor labels correctly just to the subset of rows?

Thanks! Uri

dat[is.finite(dat$z),"col"] <- cut(dat$z[is.finite(dat$z)],6)

Should work. I don't know why assigning via $ doesn't, though.

I am not totally sure that my answer is what you want, but if you want the labels instead of the integers representing the factor try as.character:

dat$col[is.finite(dat$z)] <- as.character(cut(dat$z[is.finite(dat$z)],6))

if you want it to be a factor instead of a character vector, wrap this into a call to factor:

dat$col[is.finite(dat$z)] <- factor(as.character(cut(dat$z[is.finite(dat$z)],6)))