A simulated annealing methodology for clusterwise linear regression

Wayne Desarbo, Richard L. Oliver, Arvind Rangaswamy

Research output: Contribution to journalArticle

91 Citations (Scopus)

Abstract

In many regression applications, users are often faced with difficulties due to nonlinear relationships, heterogeneous subjects, or time series which are best represented by splines. In such applications, two or more regression functions are often necessary to best summarize the underlying structure of the data. Unfortunately, in most cases, it is not known a priori which subset of observations should be approximated with which specific regression function. This paper presents a methodology which simultaneously clusters observations into a preset number of groups and estimates the corresponding regression functions' coefficients, all to optimize a common objective function. We describe the problem and discuss related procedures. A new simulated annealing-based methodology is described as well as program options to accommodate overlapping or nonoverlapping clustering, replications per subject, univariate or multivariate dependent variables, and constraints imposed on cluster membership. Extensive Monte Carlo analyses are reported which investigate the overall performance of the methodology. A consumer psychology application is provided concerning a conjoint analysis investigation of consumer satisfaction determinants. Finally, other applications and extensions of the methodology are discussed.

Original languageEnglish (US)
Pages (from-to)707-736
Number of pages30
JournalPsychometrika
Volume54
Issue number4
DOIs
StatePublished - Sep 1 1989

Fingerprint

Simulated annealing
Linear regression
Simulated Annealing
Cluster Analysis
Linear Models
Regression Function
Psychology
regression
Methodology
methodology
Conjoint Analysis
Splines
Replication
Spline
Univariate
time series
Overlapping
Time series
Determinant
psychology

All Science Journal Classification (ASJC) codes

  • Psychology(all)
  • Applied Mathematics

Cite this

Desarbo, Wayne ; Oliver, Richard L. ; Rangaswamy, Arvind. / A simulated annealing methodology for clusterwise linear regression. In: Psychometrika. 1989 ; Vol. 54, No. 4. pp. 707-736.
@article{42859eb6c91c408db394e76963bbc635,
title = "A simulated annealing methodology for clusterwise linear regression",
abstract = "In many regression applications, users are often faced with difficulties due to nonlinear relationships, heterogeneous subjects, or time series which are best represented by splines. In such applications, two or more regression functions are often necessary to best summarize the underlying structure of the data. Unfortunately, in most cases, it is not known a priori which subset of observations should be approximated with which specific regression function. This paper presents a methodology which simultaneously clusters observations into a preset number of groups and estimates the corresponding regression functions' coefficients, all to optimize a common objective function. We describe the problem and discuss related procedures. A new simulated annealing-based methodology is described as well as program options to accommodate overlapping or nonoverlapping clustering, replications per subject, univariate or multivariate dependent variables, and constraints imposed on cluster membership. Extensive Monte Carlo analyses are reported which investigate the overall performance of the methodology. A consumer psychology application is provided concerning a conjoint analysis investigation of consumer satisfaction determinants. Finally, other applications and extensions of the methodology are discussed.",
author = "Wayne Desarbo and Oliver, {Richard L.} and Arvind Rangaswamy",
year = "1989",
month = "9",
day = "1",
doi = "10.1007/BF02296405",
language = "English (US)",
volume = "54",
pages = "707--736",
journal = "Psychometrika",
issn = "0033-3123",
publisher = "Springer New York",
number = "4",

}

A simulated annealing methodology for clusterwise linear regression. / Desarbo, Wayne; Oliver, Richard L.; Rangaswamy, Arvind.

In: Psychometrika, Vol. 54, No. 4, 01.09.1989, p. 707-736.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A simulated annealing methodology for clusterwise linear regression

AU - Desarbo, Wayne

AU - Oliver, Richard L.

AU - Rangaswamy, Arvind

PY - 1989/9/1

Y1 - 1989/9/1

N2 - In many regression applications, users are often faced with difficulties due to nonlinear relationships, heterogeneous subjects, or time series which are best represented by splines. In such applications, two or more regression functions are often necessary to best summarize the underlying structure of the data. Unfortunately, in most cases, it is not known a priori which subset of observations should be approximated with which specific regression function. This paper presents a methodology which simultaneously clusters observations into a preset number of groups and estimates the corresponding regression functions' coefficients, all to optimize a common objective function. We describe the problem and discuss related procedures. A new simulated annealing-based methodology is described as well as program options to accommodate overlapping or nonoverlapping clustering, replications per subject, univariate or multivariate dependent variables, and constraints imposed on cluster membership. Extensive Monte Carlo analyses are reported which investigate the overall performance of the methodology. A consumer psychology application is provided concerning a conjoint analysis investigation of consumer satisfaction determinants. Finally, other applications and extensions of the methodology are discussed.

AB - In many regression applications, users are often faced with difficulties due to nonlinear relationships, heterogeneous subjects, or time series which are best represented by splines. In such applications, two or more regression functions are often necessary to best summarize the underlying structure of the data. Unfortunately, in most cases, it is not known a priori which subset of observations should be approximated with which specific regression function. This paper presents a methodology which simultaneously clusters observations into a preset number of groups and estimates the corresponding regression functions' coefficients, all to optimize a common objective function. We describe the problem and discuss related procedures. A new simulated annealing-based methodology is described as well as program options to accommodate overlapping or nonoverlapping clustering, replications per subject, univariate or multivariate dependent variables, and constraints imposed on cluster membership. Extensive Monte Carlo analyses are reported which investigate the overall performance of the methodology. A consumer psychology application is provided concerning a conjoint analysis investigation of consumer satisfaction determinants. Finally, other applications and extensions of the methodology are discussed.

UR - http://www.scopus.com/inward/record.url?scp=0001104737&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0001104737&partnerID=8YFLogxK

U2 - 10.1007/BF02296405

DO - 10.1007/BF02296405

M3 - Article

VL - 54

SP - 707

EP - 736

JO - Psychometrika

JF - Psychometrika

SN - 0033-3123

IS - 4

ER -