[Bioc-devel] runmean and NAs

[Valerie Obenchain]

Hi Florian,

It has been on my TODO and II've just started working on this.

On 07/10/2012 02:19 AM, Hahne, Florian wrote:
> Hi all,
> I was asking a while ago whether it would be possible for the runmean
> function to deal with NA values. Is that still on somebody's radar?
> Something similar to na.rm would be great. Haven't checked the source code
> for Rle_runsum yet, but ignoring Nas seems to be a straightforward thing
> to do. Currently I get
>
>> runmean(Rle(c(1:100, NA)), 10)
> Error in .Call2("Rle_runsum", x, as.integer(k), PACKAGE = "IRanges") :
>    some values are NA
>
>
> On a related note, why is this not working?
>> runmean(Rle(c(1:100, Inf)), 10)
> Error in .Call2("Rle_runsum", x, as.integer(k), PACKAGE = "IRanges")
> some values are NA, NaN, +/-Inf
>
>
> Shouldn't that simply return Inf for the respective vector elements?

Currently no. The Inf creates a numeric Rle

 > Rle(c(1:100, Inf))
'numeric' Rle of length 101 with 101 runs
   Lengths:   1   1   1   1   1   1   1   1 ...   1   1   1   1   1   
1   1   1
   Values :   1   2   3   4   5   6   7   8 ...  94  95  96  97  98  99 
100 Inf

so the code executed is Rle_real_runsum which has this check,

// calculate stat
if (i == 0) {
     if (!R_FINITE(*curr_value))
         error("some values are NA, NaN, +/-Inf");

While I am in there I will address this Inf issue too.

Valerie
>
> Thanks,
> Florian
>
> _______________________________________________
> Bioc-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/bioc-devel