Recursive partition and amalgamation with the exponential family

Theory and applications

A. Ciampi, Z. Lou, Qian Lin, Abdissa Negassa

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

The theory of tree‐growing (RECPAM approach) is developed for outcome variables which are distributed as the canonical exponential family. The general RECPAM approach (consisting of three steps: recursive partition, pruning and amalgamation), is reviewed. This is seen as constructing a partition with maximal information content about a parameter to be predicted, followed by simplification by the elimination of ‘negligible’ information. The measure of information is defined for an exponential family outcome as a deviance difference, and appropriate modifications of pruning and amalgamation rules are discussed. It is further shown how the proposed approach makes it possible to develop tree‐growing for situations usually treated by generalized linear models (GLIM). In particular, Poisson and logistic regression can be tree‐structured. Moreover, censored survival data can be treated, as in GLIM, by observing a formal equivalence of the likelihood under random censoring and an appropriate Poisson model. Three examples are given of application to Poisson, binary and censored survival data.

Original languageEnglish (US)
Pages (from-to)121-137
Number of pages17
JournalApplied Stochastic Models and Data Analysis
Volume7
Issue number2
DOIs
StatePublished - 1991
Externally publishedYes

Fingerprint

Censored Survival Data
Amalgamation
Exponential Family
Generalized Linear Model
Pruning
Partition
Random Censoring
Deviance
Measures of Information
Poisson Regression
Poisson Model
Information Content
Logistic Regression
Simplification
Elimination
Likelihood
Siméon Denis Poisson
Equivalence
Binary
Logistics

Keywords

  • Censored survival data
  • Generalized linear model (GLIM)
  • Logistic regression
  • Poisson regression
  • Tree growing

ASJC Scopus subject areas

  • Modeling and Simulation
  • Management of Technology and Innovation

Cite this

Recursive partition and amalgamation with the exponential family : Theory and applications. / Ciampi, A.; Lou, Z.; Lin, Qian; Negassa, Abdissa.

In: Applied Stochastic Models and Data Analysis, Vol. 7, No. 2, 1991, p. 121-137.

Research output: Contribution to journalArticle

@article{7b88c980db4144d9a28277c260ec7acc,
title = "Recursive partition and amalgamation with the exponential family: Theory and applications",
abstract = "The theory of tree‐growing (RECPAM approach) is developed for outcome variables which are distributed as the canonical exponential family. The general RECPAM approach (consisting of three steps: recursive partition, pruning and amalgamation), is reviewed. This is seen as constructing a partition with maximal information content about a parameter to be predicted, followed by simplification by the elimination of ‘negligible’ information. The measure of information is defined for an exponential family outcome as a deviance difference, and appropriate modifications of pruning and amalgamation rules are discussed. It is further shown how the proposed approach makes it possible to develop tree‐growing for situations usually treated by generalized linear models (GLIM). In particular, Poisson and logistic regression can be tree‐structured. Moreover, censored survival data can be treated, as in GLIM, by observing a formal equivalence of the likelihood under random censoring and an appropriate Poisson model. Three examples are given of application to Poisson, binary and censored survival data.",
keywords = "Censored survival data, Generalized linear model (GLIM), Logistic regression, Poisson regression, Tree growing",
author = "A. Ciampi and Z. Lou and Qian Lin and Abdissa Negassa",
year = "1991",
doi = "10.1002/asm.3150070203",
language = "English (US)",
volume = "7",
pages = "121--137",
journal = "Applied Stochastic Models in Business and Industry",
issn = "1524-1904",
publisher = "John Wiley and Sons Ltd",
number = "2",

}

TY - JOUR

T1 - Recursive partition and amalgamation with the exponential family

T2 - Theory and applications

AU - Ciampi, A.

AU - Lou, Z.

AU - Lin, Qian

AU - Negassa, Abdissa

PY - 1991

Y1 - 1991

N2 - The theory of tree‐growing (RECPAM approach) is developed for outcome variables which are distributed as the canonical exponential family. The general RECPAM approach (consisting of three steps: recursive partition, pruning and amalgamation), is reviewed. This is seen as constructing a partition with maximal information content about a parameter to be predicted, followed by simplification by the elimination of ‘negligible’ information. The measure of information is defined for an exponential family outcome as a deviance difference, and appropriate modifications of pruning and amalgamation rules are discussed. It is further shown how the proposed approach makes it possible to develop tree‐growing for situations usually treated by generalized linear models (GLIM). In particular, Poisson and logistic regression can be tree‐structured. Moreover, censored survival data can be treated, as in GLIM, by observing a formal equivalence of the likelihood under random censoring and an appropriate Poisson model. Three examples are given of application to Poisson, binary and censored survival data.

AB - The theory of tree‐growing (RECPAM approach) is developed for outcome variables which are distributed as the canonical exponential family. The general RECPAM approach (consisting of three steps: recursive partition, pruning and amalgamation), is reviewed. This is seen as constructing a partition with maximal information content about a parameter to be predicted, followed by simplification by the elimination of ‘negligible’ information. The measure of information is defined for an exponential family outcome as a deviance difference, and appropriate modifications of pruning and amalgamation rules are discussed. It is further shown how the proposed approach makes it possible to develop tree‐growing for situations usually treated by generalized linear models (GLIM). In particular, Poisson and logistic regression can be tree‐structured. Moreover, censored survival data can be treated, as in GLIM, by observing a formal equivalence of the likelihood under random censoring and an appropriate Poisson model. Three examples are given of application to Poisson, binary and censored survival data.

KW - Censored survival data

KW - Generalized linear model (GLIM)

KW - Logistic regression

KW - Poisson regression

KW - Tree growing

UR - http://www.scopus.com/inward/record.url?scp=84982402133&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84982402133&partnerID=8YFLogxK

U2 - 10.1002/asm.3150070203

DO - 10.1002/asm.3150070203

M3 - Article

VL - 7

SP - 121

EP - 137

JO - Applied Stochastic Models in Business and Industry

JF - Applied Stochastic Models in Business and Industry

SN - 1524-1904

IS - 2

ER -