Corey-- Your logic for calculating the estimated value for each cell in your design looks right to me. I believe that treating the HPD confidence intervals in the same way is correct.
While it's not exactly what you're looking for, I thought I'd provide
some related code for plotting pvals.fnc-based confidence intervals
around parameter estimates. These functions just plot each contrast
individually rather than combining them as you're attempting to do.
Still, I hope these will be useful to someone on the list.
If anyone has cleaner code or prettier plots to show off, I know I'd
appreciate seeing them.
Thanks,
/au
For this plot, you'll need to have the Design package loaded to use the
Dotplot and Cbind functions.
a.lat.p <- pvals.fnc(a.lat.l, withMCMC=TRUE)
a.lat.fixed <- a.lat.p$fixed
a.lat.lowers <- as.numeric(a.lat.fixed[2:10, 3])
a.lat.uppers <- as.numeric(a.lat.fixed[2:10, 4])
a.lat.names <- labels(a.lat.fixed)[[1]][2:10]
factors <- as.factor(a.lat.names)
Dotplot(factors ~ Cbind(a.lat.estimates, a.lat.lowers, a.lat.uppers),
data = a.lat.p$fixed,
ylab = NULL, xlab = "change in latency (ms)",
panel = function (x, y, ...) {
panel.Dotplot(x,y, ...)
panel.abline(v=0)
})
dev.off()
And here's some code for more traditional bar plots, again based on
pvals.fnc.
lat.p <- pvals.fnc(lat.m, withMCMC=TRUE)
lat.estimates <- as.numeric(as.matrix(lat.p$fixed[2:4, 2]))
lat.lowers <- as.numeric(lat.p$fixed[2:4, 3])
lat.uppers <- as.numeric(lat.p$fixed[2:4, 4])
lat.param.plot <- barplot(lat.estimates,
beside=TRUE,
ylim=c(min(lat.lowers)-1, max(lat.uppers)+1),
ylab="Change in latency (ms)")
segments(x0=lat.param.plot,
x1=lat.param.plot,
y0=lat.lowers,
y1=lat.uppers)
segments(x0=lat.param.plot-.1,
x1=lat.param.plot+.1,
y0=lat.lowers,
y1=lat.lowers)
segments(x0=lat.param.plot-.1,
x1=lat.param.plot+.1,
y0=lat.uppers,
y1=lat.uppers)
abline(0,0)
pgp0ruDZhcG3y.pgp
Description: PGP signature
_______________________________________________ R-lang mailing list [email protected] http://pidgin.ucsd.edu/mailman/listinfo/r-lang
