Right now it only works with zero values. Which is silly. Missing should be either zero or NA, or one or the other. I would use the same trick we use in `ici_kendalltau`, where we can create a vector of missing values.