Rfast::colMads() crashing R session on NA

[thewoodsofcoding]

Hi, I use Rfast for robust calculations on huge data sets, say 1000+ rows for each column. I want to calculate col medians and col mads. It happens that Rfast::colMedians() and Rfast::colMads() have no issue with NA but only calculate median or mad if NA is <50% of the columns content, however I want it also to be calculated even if there are only, lets say 2 of 1000 values != NA, so I use na.rm = T. I also have columns that have only NA and thous will be calculated as NA for Rfast::colMedians(mat, na.rm = T) so it works fine. But for Rfast::colMads(mat, na.rm = T) R session crashes as soon there are NA only columns in mat. Sure I can remove NA only columns prior to calculation, but I think this is a bug of Rfast::colMads(mat, na.rm = T) function. Attached find an example for reproduction:

noNA <- c(1,2,3,4)
someNA <- c(1,2,3,NA)
equalNA <- c(1,2,NA,NA)
moreNA <- c(1,NA,NA,NA)
onlyNA <- c(NA,NA,NA,NA)
mat <- cbind(noNA, someNA, equalNA, moreNA, onlyNA)
mat
#       noNA someNA equalNA moreNA onlyNA
# [1,]    1      1       1      1     NA
# [2,]    2      2       2     NA     NA
# [3,]    3      3      NA     NA     NA
# [4,]    4     NA      NA     NA     NA
Rfast::colMedians(mat)
# [1] 2.5 2.5  NA  NA  NA    #too much NA here
Rfast::colMedians(mat, na.rm = T)
# [1] 2.5 2.0 1.5 1.0  NA    #desired result
Rfast::colMads(mat)
# [1] 1.4826 1.4826     NA     NA     NA
Rfast::colMads(mat, na.rm = T)
# R session crashes

[ManosPapadakis95]

It is not a bug. It's just that I have never thought that someone would really have this case in his data. What do you want as result?

[thewoodsofcoding]

Thanks for the fast response! Best would be to return an NA for cases of only one value or non value (only NA) columns.

[ManosPapadakis95]

Done.