spatialSign NA handling [feature suggestion]

[unitroot]

Although imputation is the methodically sound way to go in most cases, sometimes I find myself performing imputations at the very end of pre-processing (mostly with time series). Therefore, it would be nice to have a NA remove feature in the spatialSign function:

"spatialSign.default" <- function(x) { if(is.character(x) | is.factor(x)) stop("spatial sign is not defined for character or factor data") denom <- sum(x^2, na.rm = TRUE) out <- if(sqrt(denom) > .Machine$double.eps) x / sqrt(denom) else x * 0 out }

Conceptually, this is equivalent to treating NAs as zeros, which is very often at least not totally wrong.

[topepo]

Good point! See 4cc54a3.