On Fri, 29 Jan 1999, Barker, Chris {GL P~Palo Alto} wrote:
>
> I have frequently used RPART without problems. However, I've started to
> encounter a problem with RPART detecting NAN's in my data even though
> I don't appear to have any in the variables I'm using. Any suggestions on
> "debugging" my data appreciated.
>
> > quick <- data.frame(bpd24, wt.4, bpd1, bpd12)
> > summary(quick)
> bpd24 wt.4 bpd1 bpd12
> Min.: 41.00 Min.: 61.00 Min.: 48.00 Min.: 48.00
> 1st Qu.: 70.00 1st Qu.: 88.80 1st Qu.: 72.00 1st Qu.: 71.00
> Median: 80.00 Median: 98.50 Median: 80.00 Median: 80.00
> Mean: 78.47 Mean: 99.58 Mean: 79.05 Mean: 78.71
> 3rd Qu.: 85.00 3rd Qu.:109.10 3rd Qu.: 85.00 3rd Qu.: 85.00
> Max.:120.00 Max.:155.50 Max.:120.00 Max.:120.00
> NA's:710.00 NA's: 20.00 NA's: 30.00 NA's:403.00
> > fit.bpd <- rpart(bpd24 ~ wt.4 + bpd1 + bpd12, control = rpart.control(cp =
>
> 0.005))
> Error in .C("s_to_rp",: subroutine s_to_rp: 126 NaN value(s) in argument 10
> Dumped
The usual story is that those NAs are really NaNs. You can't tell at the
S level, but the .C interface can, and it objects to NaNs unless told
not to.
Try
bpd24[is.na(bpd24)] <- NA
etc and see if you see any difference.
You can probably add specialsok=T to the .C("s_to_rp" call, but
don't rely on this, as I have not tested it.
--
Brian D. Ripley, ripley@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 272860 (secr)
Oxford OX1 3TG, UK Fax: +44 1865 272595
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
|