If a variable is completely NA then `ds.mean()` and `ds.var()` return the error: ``` Nvalid less than nfilter.tab ``` and the call fails. I am not sure that is the right behaviour because that is confusing for the analyst, perhaps they should return NA and not fail?