R devel - Mar 2001 - histogram of frequencies (PR#861)

Full_Name: Dr. Bernd Winkler
Version: 1.2.2
OS: Linux
Submission from: (NULL) (194.59.179.176)


Hi all,

having updated to version 1.2.2 last week I experienced some strange behaviour
of 
the hist function.

I want to plot a histogram of frequencies resp. counts, for example poisson 
random numbers, on a predefined and fixed x scale
> hist(rpois(500, 2), breaks=0:10)
But instead of the frequencies I get the density of these data.
Introducing the parameter freq=TRUE then I get the frequencies but 
with the following warning message
> hist(rpois(500, 2),breaks=0:10, freq=TRUE)
Warning message: 
the AREAS in the plot are wrong -- rather use `freq=FALSE'! in:
plot.histogram(r, freq = freq, col = col, border = border, main = main,  

The following excerpt from hist.default(...) (new in 1.2.2) might have caused
this behaviour:
    ...
    diddle <- 1e-07 * max(abs(range(breaks)))
    fuzz <- if (right) 
        c(if (include.lowest) -diddle else diddle, rep(diddle, 
            length(breaks) - 1))
    else c(rep(-diddle, length(breaks) - 1), if (include.lowest) diddle else
-diddle)
    breaks <- breaks + fuzz
    ...

The addition of the fuzz variable to the breaks causes hist 
to think of no equidistant breaks (even if they are) and therefore 
a density plot is generated by default. Removing these lines the 
behaviour is like the one in version 1.2.1 and earlier. 
Is there an alternative doing this with the new version?
And what about the border color parameter in barplot?

Many thanks in advance

Bernd





-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

bernd.winkler@genedata.com writes:
> The addition of the fuzz variable to the breaks causes hist 
> to think of no equidistant breaks (even if they are) and therefore 
> a density plot is generated by default. Removing these lines the 
> behaviour is like the one in version 1.2.1 and earlier. 
> Is there an alternative doing this with the new version?
> And what about the border color parameter in barplot?
Argh! This just goes to prove how hard it is to fix one bug without
introducing another.... 

The two types of fuzz in hist.default should not be of the same size! 
Try setting the multiplier in "diddle" to a smaller value, like 1e-12.
Either that or we should test for equidistant breaks before adding the
fuzz. 

It only happens with user-specified breaks, though, and the error
message with freq=T is of course ignorable.

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._