Any #rstats people know why the sd(x) function in base R doesn't work but sapply(x, sd) does? Is everyone just piping things into the sd function these days? Is there a secret sd function club I'm not a member of? Seems kind of an important, basic stats operation to not work.

Provide a MCVE or it doesn't count :) > sd(1:10) [1] 3.02765 > set.seed(123); sd(rnorm(10)) [1] 0.953784 > Is you 'x' a list or data.frame? Well yes in that case you need to loop over columns.

A dataframe. > sd(shark) Error in is.data.frame(x) : 'list' object cannot be coerced to type 'double' > sapply(shark,sd) Length 2.549928

Yup, rings a bell. Might be historical compatibility. There are add-on packages with column-wise variance or sd. Because this is a cludge -- but been-there-done-that: > library(palmerpenguins) > sapply(penguins, function(x) if (is.numeric(x)) sd(na.omit(x)) else NULL)

Doing mean() and sd() to whole data frames got removed in 3.0.0

Did they just used to error in the (pretty frequent) cases where doing so doesn't make sense?

I don't remember -- it might have returned NA for things that weren't obviously numeric