R help - Sep 2005 - Problem with memory footprint of qq plot generated with lattice

Dear R helpers,

I generate a qq plot using the following function call.

qqmath(~val|ind,data=xx
       ,distribution=function(p) qt(p,df=19)
       ,ylab="Sample Quatinles"
       ,xlab="Theoretical Quantiles"
       ,aspect=1
       ,prepanel = prepanel.qqmathline
       ,panel=function(x,y)
       {
         panel.qqmathline(y, distribution=function(p) qt(p,df=19),col=2)
         panel.qqmath(x, y , distribution=function(p) 
qt(p,df=19),pch=".",cex=2)
       }
)

dim(xx)
[1] 680237      2

unique(xx[,1])
[1] 500-530   1000-1110 2000-2200 3400-3700
Levels: 500-530 1000-1110 2000-2200 3400-3700

It means that actually four small qqplots are drawn.

The problem is that the generated pdf is extremely large.

-rw-r--r--  1 nwew bayes 19734746 2005-08-27 15:38 Distr-qqplot.pdf

Suggestions howe to reduce the size but keep the quality are wellcome.

cheers
Eryk

nwew <W.E.Wolski at newcastle.ac.uk> wrote:
> Dear R helpers,
> I generate a qq plot using the following function call.
...
> dim(xx)
> [1] 680237      2
How about doing something like this:

fn <- function(n,cut=0.001,m=1000)
{
    p <- ppoints(n)
    p <- p[pmin(p, 1-p) < cut]
    q <- pt(seq(qt(cut,df=19),qt(1-cut,df=19),length=m),df=19)
    sort(c(p,q))
}

then adding 'f.value=fn' to your qqmath arguments?  This essentially
says, plot the individual data points in the extreme tails of the
distribution (p < 0.001 or p > 0.999), and evaluate the distribution
at a sparse set of points in between, where the density means you
can't discern the individual values anyway.

-- Dave