[R] More basic equivalent of TukeyHSD

Janet Rosenbaum janet.rosenbaum at gmail.com
Fri Mar 13 22:43:10 CET 2009


This is a simple question, but I'm going on the supposition that the
only stupid question is the one not asked.

1.  I have many sets of 5 proportions that are different from each
other (prop.test), and want to know which proportions are different
from each other.  In other words, I want the equivalent of Tukey's HSD
test, but for proportions rather than anova.

Here is one of the sets of number of "successes"; the denominator is
151 for all cases.
                   Parents School Peers Church No one
Menstruation             79     58    55      3     38

As proportions:
> round(topics.sum[1,], digits=3)
Parents  School   Peers  Church  No one
 0.523   0.384   0.364   0.020   0.252

Intuitively, it looks like there are 3 or 4 groups that are different
from each other, but is there a function that does this?
parents > school and peers ?> no one > church

2.  I am displaying this data in a bar graph.  How do I show
graphically which proportions are different?  I'm not interested in
using error bars since they can be ambiguous, but probably something
like putting a different symbol over each bar to show which bars are
the same and which differ.  I've made efforts at this before, but
they've always looked a bit clumsy.

Thanks,

Janet




More information about the R-help mailing list