Quantile transform sometimes generates NA values

[rplzzz]

The problem seems to be with the interpolation function used to approximate the quantile function:

fldgen/R/normalizeresiduals.R

Lines 176 to 187 in d7922a4

	inv.ecdf <- function(column.resids) {
	# for one column of the residuals, get the order of points.
	order.vector <- rank(column.resids)
	# cumulative probabilities are order-1 / Ntime+1
	probabilities <- (order.vector - 1) * offset
	# return the ecdf for the function
	approxfun(x = probabilities, y = column.resids)
	}

By default, approxfun produces NA values for inputs outside the range of x-values used to construct the approximation. In theory, this should never happen, but due to roundoff error the largest or smallest value in the empirical distribution can sometimes appear to be out of range. The easiest fix will be to specify rule=2 in the call to approxfun, which will cause these values to be clamped to the range of the x-values used to construct the approximator, which is what we want anyhow.