[R] Odd behaviour of mean() with a numeric column in a tibble

[Ista Zahn]

Not at a computer to check right now, but I believe single bracket indexing
a tibble always returns a tibble. To extract a vector use [[

On Dec 6, 2016 4:28 PM, "Chris Evans" <chrishold at psyctc.org> wrote:
>
> I hope I am obeying the list rules here. I am using a raw R IDE for this
and running 3.3.2 (2016-10-31) on x86_64-w64-mingw32/x64 (64-bit)
>
> Here is a reproducible example.  Code only first
>
> require(tibble)
> tmpTibble <- tibble(ID=letters,num=1:26)
> min(tmpTibble[,2]) # fine
> max(tmpTibble[,2]) # fine
> median(tmpTibble[,2])  # not fine
> mean(tmpTibble[,2])    # not fine

I think you want

mean(tmpTibble[[2]]

> newMeanFun <- function(x) {mean(as.numeric(unlist(x)))}
> newMeanFun(tmpTibble[,2]) # solved problem but surely shouldn't be
necessary?!
> newMedianFun <- function(x) {median(as.numeric(unlist(x)))}
> newMedianFun(tmpTibble[,2]) # ditto
> str(tmpTibble[,2])
>
> ### then I tried this to make sure it wasn't about having fed in integers
>
> tmpTibble2 <- tibble(ID=letters,num=1:26,num2=(1:26)/10)
> tmpTibble2
> mean(tmpTibble2[,3]) # not fine, not about integers!
>
>
> ### before I just created tmpTibble2 I found myself trying to add a
column to tmpTibble
> tmpTibble$newNum <- tmpTibble[,2]/10  # NO!
> tmpTibble[["newNum"]] <- tmpTibble[,2]/10 # NO!
> ### and oddly enough ...
> add_column(tmpTibble,newNum = tmpTibble[,2]/10) # NO!
>
> Now here it is with the output:
>
> > require(tibble)
> Loading required package: tibble
> > tmpTibble <- tibble(ID=letters,num=1:26)
> > min(tmpTibble[,2]) # fine
> [1] 1
> > max(tmpTibble[,2]) # fine
> [1] 26
> > median(tmpTibble[,2])  # not fine
> Error in median.default(tmpTibble[, 2]) : need numeric data
> > mean(tmpTibble[,2])    # not fine
> [1] NA
> Warning message:
> In mean.default(tmpTibble[, 2]) :
>   argument is not numeric or logical: returning NA
> > newMeanFun <- function(x) {mean(as.numeric(unlist(x)))}
> > newMeanFun(tmpTibble[,2]) # solved problem but surely shouldn't be
necessary?!
> [1] 13.5
> > newMedianFun <- function(x) {median(as.numeric(unlist(x)))}
> > newMedianFun(tmpTibble[,2]) # ditto
> [1] 13.5
> > str(tmpTibble[,2])
> Classes ‘tbl_df’, ‘tbl’ and 'data.frame':       26 obs. of  1 variable:
>  $ num: int  1 2 3 4 5 6 7 8 9 10 ...
> >
> > ### then I tried this to make sure it wasn't about having fed in
integers
> >
> > tmpTibble2 <- tibble(ID=letters,num=1:26,num2=(1:26)/10)
> > tmpTibble2
> # A tibble: 26 × 3
>       ID   num  num2
>    <chr> <int> <dbl>
> 1      a     1   0.1
> 2      b     2   0.2
> 3      c     3   0.3
> 4      d     4   0.4
> 5      e     5   0.5
> 6      f     6   0.6
> 7      g     7   0.7
> 8      h     8   0.8
> 9      i     9   0.9
> 10     j    10   1.0
> # ... with 16 more rows
> > mean(tmpTibble2[,3]) # not fine, not about integers!
> [1] NA
> Warning message:
> In mean.default(tmpTibble2[, 3]) :
>   argument is not numeric or logical: returning NA
> >
> >
> > ### before I just created tmpTibble2 I found myself trying to add a
column to tmpTibble
> > tmpTibble$newNum <- tmpTibble[,2]/10  # NO!
> > tmpTibble[["newNum"]] <- tmpTibble[,2]/10 # NO!
> > ### and oddly enough ...
> > add_column(tmpTibble,newNum = tmpTibble[,2]/10) # NO!
> Error: Each variable must be a 1d atomic vector or list.
> Problem variables: 'newNum'
> >
> >
>
> I discovered this when I hit odd behaviour after using read_spss() from
the haven package for the first time as it seemed to be offering a step
forward over good old read.spss() from the excellent foreign package.  I am
reporting it here not directly to Prof. Wickham as the issues seem rather
general though I'm guessing that it needs to be fixed with a fix to
tibble.   Or perhaps I've completely missed something.
>
> TIA,
>
> Chris
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

	[[alternative HTML version deleted]]