I would aprreciate your help with regard to the following issue
(meanwhile you can pick up a little trick for displaying confidence
limits):
The data frame EXMPL1 contains the midpoints and the lower and
upper limits of confidence intervals for two measurement types
and the mean differences B-A and C-A
EXMPL1
M.TYPE DIFF ESTIMATE PVALUE
1 D B-A 8.4 p=0.04
2 D B-A 9.0 p=0.04
3 D B-A 9.6 p=0.04
4 D C-A 12.4 p=0.03
5 D C-A 13.2 p=0.03
6 D C-A 14.0 p=0.03
7 S B-A 18.4 p=0.01
8 S B-A 19.0 p=0.01
9 S B-A 19.6 p=0.01
10 S C-A 22.4 p=0.5
11 S C-A 23.2 p=0.5
12 S C-A 24.0 p=0.5
We define the panel function panel.bwconfi by replacing the line
q <- quantile(X, c(0.75, 0.5, 0.25)) in panel.bwplot through
q <- quantile(X, c(1, 0.5, 0)).
This little trick helps us to plot the confidence intervals
as boxes in a trellis graphics:
bwplot(DIFF~ESTIMATE | M.TYPE, data=EXMPL1, panel=function(x,y)
{panel.bwconfi(x,y); layout=c(1,2))
But now we want to add the corresponding p-values (this does not make to
much sense,
but some people like p-values more than they should):
bwplot(DIFF~ESTIMATE | M.TYPE, data=EXMPL1, panel=function(x,y)
{panel.bwconfi(x,y); text(16,y,EXMPL1$PVALUE)}, layout=c(1,2))
This doesn't work because the panel function doesn't see the levels
of M.TYPE. The different p-values are overlaid.
How can I get the plot with p-values?
More generally: How can I let depend the panel function on the levels
of the conditioning factors?
Kind regards
Prof. Juergen Bock, F.Hoffmann-La Roche, Basel, Switzerland
-----------------------------------------------------------------------
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