In our Summer Stats Institute, I was asked a question that amounts to
reversing the effect of the contrasts function (reconstruct an ordinal
predictor from a set of binary columns). The best I could think of was to
link together several ifelse functions, and I don't think I want to do this
if the example became any more complicated.
I'm unable to remember a less error prone method :). But I expect you might.
Here's my working example code
## Paul Johnson <pauljohn at ku.edu>
## 2013-06-07
## We need to create an ordinal factor from these indicators
## completed elementary school
es <- c(0, 0, 1, 0, 1, 0, 1, 1)
## completed high school
hs <- c(0, 0, 1, 0, 1, 0, 1, 0)
## completed college graduate
cg <- c(0, 0, 0, 0, 1, 0, 1, 0)
ed <- ifelse(cg == 1, 3,
ifelse(hs == 1, 2,
ifelse(es == 1, 1, 0)))
edf <- factor(ed, levels = 0:3, labels = c("none", "es", "hs", "cg"))
data.frame(es, hs, cg, ed, edf)
## Looks OK, but what if there are missings?
es <- c(0, 0, 1, 0, 1, 0, 1, 1, NA, NA)
hs <- c(0, 0, 1, 0, 1, 0, 1, 0, 1, NA)
cg <- c(0, 0, 0, 0, 1, 0, 1, 0, NA, NA)
ed <- ifelse(cg == 1, 3,
ifelse(hs == 1, 2,
ifelse(es == 1, 1, 0)))
cbind(es, hs, cg, ed)
## That's bad, ifelse returns NA too frequently.
## Revise (becoming tedious!)
ed <- ifelse(!is.na(cg) & cg == 1, 3,
ifelse(!is.na(hs) & hs == 1, 2,
ifelse(!is.na(es) & es == 1, 1,
ifelse(is.na(es), NA, 0))))
cbind(es, hs, cg, ed)
## Does the project director want us to worry about
## logical inconsistencies, such as es = 0 but cg = 1?
## I hope not.
Thanks in advance, I hope you are having a nice summer.
pj
--
Paul E. Johnson
Professor, Political Science Assoc. Director
1541 Lilac Lane, Room 504 Center for Research Methods
University of Kansas University of Kansas
http://pj.freefaculty.org http://quant.ku.edu
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
