24 lines
928 B
R
24 lines
928 B
R
library(rpart)
|
|
mystate <- data.frame(state.x77, region=factor(state.region))
|
|
names(mystate) <- c("population","income" , "illiteracy","life" ,
|
|
"murder", "hs.grad", "frost", "area", "region")
|
|
#
|
|
# This little test came out of a query that cp did not scale
|
|
# with the data size. It does
|
|
#
|
|
# tdata = 20 copies of "mystate"
|
|
# trees with tdata and trees with mystate should be the same (they are)
|
|
# except for the n's
|
|
set.seed(10)
|
|
|
|
tdata <- rbind(mystate, mystate, mystate, mystate, mystate)
|
|
tdata <- rbind(tdata, tdata, tdata, tdata)
|
|
tfit1 <- rpart(income ~ population + illiteracy + murder + hs.grad + region,
|
|
data = mystate, method = "anova", xval=0, cp=.089)
|
|
tfit2 <- rpart(income ~ population + illiteracy + murder + hs.grad + region,
|
|
data = tdata, method='anova', xval=0, cp=.089,
|
|
minsplit=400, minbucket=140)
|
|
|
|
all.equal(tfit1$splits[,-1], tfit2$splits[,-1])
|
|
|