[R] rpart- error rate for survival regression?
Hyung Kim
spok_1 at yahoo.com
Tue Feb 11 15:42:32 CET 2003
Hello
I am using rpart with survival data. When I 'printcp'
I get the results below. Can someone please explain
two things for me in lay man's terms. 1) What exactly
is the complexity parameter? 2) How do I calculate
the absolute cross validated error rate when the root
node error in my example is 1.39?
Thanks you.
spok leen
Survival regression tree:
rpart(formula = Surv(fu, death) ~ p + k + c +
g + a + e, data = dat2, xval = 700)
Variables actually used in tree construction:
[1] c e k p a
Root node error: 972.86/700 = 1.3898
n= 700
CP nsplit rel error xerror xstd
1 0.068408 0 1.00000 1.00390 0.048144
2 0.036091 1 0.93159 0.93917 0.048700
3 0.025606 2 0.89550 0.94267 0.052853
4 0.022596 3 0.86990 0.90079 0.052251
5 0.011670 4 0.84730 0.87244 0.050768
6 0.010000 5 0.83563 0.90675 0.053937
More information about the R-help
mailing list