R help - Jan 2005 - bwplot: how not to draw outliers

RenE J.V. Bertin wrote:
> Hello, and (somewhat belated) best wishes for 2005.
> 
> Can one order not to draw outliers in bwplot, or at least exclude them from
the vertical axis scaling? If so, how (or what doc do I need to consult)?
> The options that have this effect in boxplot() do not appear to have any
effect with bwplot (although outline=FALSE in boxplot does *not* change the
scaling).
> 
> Thanks,
> RenE Bertin
> 

RenE,

There may be other solutions but you can do this using the prepanel 
option to set the ylim:

library(lattice)
set.seed(1)
z <- data.frame(x = rt(100, 1), g = rep(letters[1:4], each = 25))
bwplot(x ~ g, z,
        prepanel = function(x, y) {
          bp <- boxplot(split(y, x), plot = FALSE)
          ylim <- range(bp$stats)
          list(ylim = ylim)
        })

If you actually want to exclude the points (rather than just prevent 
outliers from affecting the scale), you will have to modify the 
panel.bwplot function in addition to using the above.

--sundar

On Monday 17 January 2005 14:51, Sundar Dorai-Raj wrote:
> RenE J.V. Bertin wrote:
> > Hello, and (somewhat belated) best wishes for 2005.
> >
> > Can one order not to draw outliers in bwplot, or at least exclude
> > them from the vertical axis scaling? If so, how (or what doc do I
> > need to consult)? The options that have this effect in boxplot() do
> > not appear to have any effect with bwplot (although outline=FALSE
> > in boxplot does *not* change the scaling).
> >
> > Thanks,
> > RenE Bertin
>
> RenE,
>
> There may be other solutions but you can do this using the prepanel
> option to set the ylim:
>
> library(lattice)
> set.seed(1)
> z <- data.frame(x = rt(100, 1), g = rep(letters[1:4], each = 25))
> bwplot(x ~ g, z,
>         prepanel = function(x, y) {
>           bp <- boxplot(split(y, x), plot = FALSE)
>           ylim <- range(bp$stats)
>           list(ylim = ylim)
>         })
>
> If you actually want to exclude the points (rather than just prevent
> outliers from affecting the scale), you will have to modify the
> panel.bwplot function in addition to using the above.
Right. panel.bwplot calls 

       stats <- boxplot.stats(y[x == xval], coef = coef)

I guess it should have a 'do.out' argument as well. A workaround 
(without changing the panel function) might be to use something like

bwplot(voice.part ~ height, data = singer, 
       par.settings = list(plot.symbol = list(col = "transparent")))

(neither of these would change the limits, of course)

-Deepayan

Sundar Dorai-Raj wrote:
> 
> 
> RenE J.V. Bertin wrote:
> 
>> Hello, and (somewhat belated) best wishes for 2005.
>>
>> Can one order not to draw outliers in bwplot, or at least exclude them 
>> from the vertical axis scaling? If so, how (or what doc do I need to 
>> consult)?
>> The options that have this effect in boxplot() do not appear to have 
>> any effect with bwplot (although outline=FALSE in boxplot does *not* 
>> change the scaling).
>>
>> Thanks,
>> RenE Bertin
>>
> 
> 
> RenE,
> 
> There may be other solutions but you can do this using the prepanel 
> option to set the ylim:
> 
> library(lattice)
> set.seed(1)
> z <- data.frame(x = rt(100, 1), g = rep(letters[1:4], each = 25))
> bwplot(x ~ g, z,
>        prepanel = function(x, y) {
>          bp <- boxplot(split(y, x), plot = FALSE)
>          ylim <- range(bp$stats)
>          list(ylim = ylim)
>        })
> 
> If you actually want to exclude the points (rather than just prevent 
> outliers from affecting the scale), you will have to modify the 
> panel.bwplot function in addition to using the above.
> 
> --sundar
You may also want to try

library(Hmisc)
library(lattice)
bwplot(..., panel=panel.bpplot)
?panel.bpplot


-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University