lrpPVal {longRPart} | R Documentation |
Using permutation tests, this function calculates the p-value for the first split of the tree; see the linked pdf for details. It also calculates the p-value for the entire tree, though the statistical theory behind this p-value is not as clear.
lrpPVal(model,J)
model |
model produced by longRPart |
J |
number of permutations to run |
The basic principle behind the test is to randomized the covariates to the subject and then recalculate the tree: the p-value is the proportion of trees that have a greater reduction in deviance than the original tree.
J indicates that a longRPart tree will be built J times, and a reasonable value for J would be 100 (at least), which should take 100 times as long as the original longRPart function
Mohamed Abdolell <mo@dal.ca> and Sam Stewart
Abdolell et al. Binary partitioning for continuous longitudinal data: categorizing a prognostic variable. Statistics in Medicine (2002) vol. 21 pp. 3395-3409