I'm working on an example of plotting predicted probabilities from a
proportional odds model.
The steps below generate a data frame, plotdat, that I want to plot with
ggplot.
library(MASS)
data("Arthritis", package="vcd")
arth.polr <- polr(Improved ~ Sex + Treatment + Age, data=Arthritis,
Hess=TRUE)
# get predicted probs for categories of Improve
arth.fitp <- cbind(Arthritis,
predict(arth.polr, type="probs"))
head(arth.fitp)
# reshape probs to long
library(reshape2)
plotdat <- melt(arth.fitp,
id.vars = c("Sex", "Treatment",
"Age", "Improved"),
measure.vars=c("None", "Some",
"Marked"),
variable.name = "Level",
value.name = "Probability")
## view first few rows
head(plotdat)
> head(plotdat)
Sex Treatment Age Improved Level Probability
1 Male Treated 27 Some None 0.7326185
2 Male Treated 29 None None 0.7174048
3 Male Treated 30 None None 0.7096042
4 Male Treated 32 Marked None 0.6936286
5 Male Treated 46 Marked None 0.5702499
6 Male Treated 58 Marked None 0.4563432
In the plot step, I am plotting Probability vs. Age, stratified by
Level, and faceted by
Sex and Treatment. My question concerns the use of geom_point().
The call below plots 3 points for each case, one on each Level curve.
ggplot(plotdat, aes(x = Age, y = Probability, color = Level)) +
geom_line(size=2.5) + theme_bw() + xlim(10,80) +
geom_point(color="black", size=1.5) +
facet_grid(Sex ~ Treatment,
labeller = function(x, y) sprintf("%s = %s", x, y)
)
Instead,
I want to plot only one point for each case, for the value of Level that
corresponds
to the value of Improved in this data set. Somehow, this involves something
like an aes() argument to geom_point(), with Level indexed by Improved,
or some such.
How can I do this?
--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA