Pruning regression trees for censored survival data: the recpam approach |
| |
Authors: | Antonio Ciampi Johanne Thiffault |
| |
Affiliation: | Montreal Childrens Hospital Research Institute , 4060 rue Ste-Catherine ouest, Montreal Que, H3Z 2Z3, Canada |
| |
Abstract: | RECPAM is a methodology, implemented in a computer program of the same name, for the construction of tree-structured models in Biostatistics. In this work we present algorithms for pruning and amalgamating terminal nodes of a tree, within the RECPAM approach. These algorithms construct sequences of nested models and calculate at each step the AIC of the corresponding model and correct significance levels, according to Gabriel's theory of Simultaneous Test Procedures. As an example, the analysis of data from clinical trials involving patients with Small Cell Carcinoma of the Lung is presented. |
| |
Keywords: | survival analysis multiple comparisons significance levels of regression trees Akaike Information Criterion |
|
|