R devel - Jan 2011 - naresid.exclude query

x <- NA
  na.act <- na.action(na.exclude(x))
  y <- rep(0,0)
  naresid(na.act,y)

... currently produces the result...
  numeric(0)

... whereas the documentation might lead you to expect
  NA

The behaviour is caused by the line 
  if (length(x) == 0L) return(x)

in `stats:::naresid.exclude'. Removing this line results in the behaviour
I'd
expected in the above example (and in a test example where `x' is a zero row
matrix). 

Is the coded behaviour necessary for some reason? Could it be changed (so that 
my example produces NA)? The reason I ask is that I use `napredict' in 
mgcv:predict.gam, and someone complained that if he predicts using newdata 
that is all NA, then he doesn't get what he expected (he has a pretty good 
reason for doing this). Part of the problem with predict.gam in this case was 
my code, but once I fixed that I ran up against the above problem.  

Actually, I just checked the source code and the line of naresid.exclude that 
causes the problem already has this comment after it....
  # << FIXME? -- reconstructing all NA object
... so I guess I'm really asking if there is any chance of fixing this soon,
or whether I should just code up a work-around for predict.gam?

Simon

-- 
> Simon Wood, Mathematical Sciences, University of Bath, Bath, BA2 7AY UK
> +44 1225 386603  www.maths.bath.ac.uk/~sw283

There was a reason for that line, as running 'make check' with your 
modification would have shown you.  Sometimes naresid() is called with 
x=NULL.   AFAICS replacing the test with is.null(x) suffices, and I'm 
testing that in R-devel.  If nothing else comes up I will port it to 
R-patched later.

Note that R's own code does not AFAIR call these functions when 
predicting with 'newdata', and a fit with all cases omitted would not 
be very interesting so it is unsurprising no one noticed for >8 years.

On Fri, 14 Jan 2011, Simon Wood wrote:
>  x <- NA
>  na.act <- na.action(na.exclude(x))
>  y <- rep(0,0)
>  naresid(na.act,y)
>
> ... currently produces the result...
>  numeric(0)
>
> ... whereas the documentation might lead you to expect
>  NA
>
> The behaviour is caused by the line
>  if (length(x) == 0L) return(x)
>
> in `stats:::naresid.exclude'. Removing this line results in the
behaviour I'd
> expected in the above example (and in a test example where `x' is a
zero row
> matrix).
>
> Is the coded behaviour necessary for some reason? Could it be changed (so
that
> my example produces NA)? The reason I ask is that I use `napredict' in
> mgcv:predict.gam, and someone complained that if he predicts using newdata
> that is all NA, then he doesn't get what he expected (he has a pretty
good
> reason for doing this). Part of the problem with predict.gam in this case
was
> my code, but once I fixed that I ran up against the above problem.
>
> Actually, I just checked the source code and the line of naresid.exclude
that
> causes the problem already has this comment after it....
>  # << FIXME? -- reconstructing all NA object
> ... so I guess I'm really asking if there is any chance of fixing this
soon,
> or whether I should just code up a work-around for predict.gam?
>
> Simon
>
> -- 
>> Simon Wood, Mathematical Sciences, University of Bath, Bath, BA2 7AY UK
>> +44 1225 386603  www.maths.bath.ac.uk/~sw283
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595