Clustering and variable selection in the presence of mixed variable types and missing data

Curtis Storlie, S. M. Myers, Slavica K Katusic, A. L. Weaver, R. G. Voigt, Paul E Croarkin, R. E. Stoeckel, John D Port

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

We consider the problem of model-based clustering in the presence of many correlated, mixed continuous, and discrete variables, some of which may have missing values. Discrete variables are treated with a latent continuous variable approach, and the Dirichlet process is used to construct a mixture model with an unknown number of components. Variable selection is also performed to identify the variables that are most influential for determining cluster membership. The work is motivated by the need to cluster patients thought to potentially have autism spectrum disorder on the basis of many cognitive and/or behavioral test scores. There are a modest number of patients (486) in the data set along with many (55) test score variables (many of which are discrete valued and/or missing). The goal of the work is to (1) cluster these patients into similar groups to help identify those with similar clinical presentation and (2) identify a sparse subset of tests that inform the clusters in order to eliminate unnecessary testing. The proposed approach compares very favorably with other methods via simulation of problems of this type. The results of the autism spectrum disorder analysis suggested 3 clusters to be most likely, while only 4 test scores had high (>0.5) posterior probability of being informative. This will result in much more efficient and informative testing. The need to cluster observations on the basis of many correlated, continuous/discrete variables with missing values is a common problem in the health sciences as well as in many other disciplines.

Original languageEnglish (US)
Pages (from-to)2884-2899
Number of pages16
JournalStatistics in Medicine
Volume37
Issue number19
DOIs
StatePublished - Aug 30 2018

Fingerprint

Variable Selection
Missing Data
Cluster Analysis
Clustering
Discrete Variables
Score Test
Continuous Variables
Missing Values
Disorder
Spectrum Analysis
Model-based Clustering
Testing
Dirichlet Process
Health
Posterior Probability
Number of Components
Latent Variables
Mixture Model
Simulation Methods
Eliminate

Keywords

  • Dirichlet process
  • hierarchical Bayesian modeling
  • missing data
  • mixed variable types
  • model-based clustering
  • variable selection

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Cite this

Clustering and variable selection in the presence of mixed variable types and missing data. / Storlie, Curtis; Myers, S. M.; Katusic, Slavica K; Weaver, A. L.; Voigt, R. G.; Croarkin, Paul E; Stoeckel, R. E.; Port, John D.

In: Statistics in Medicine, Vol. 37, No. 19, 30.08.2018, p. 2884-2899.

Research output: Contribution to journalArticle

@article{d609cbdf206d4189a373796305136ac7,
title = "Clustering and variable selection in the presence of mixed variable types and missing data",
abstract = "We consider the problem of model-based clustering in the presence of many correlated, mixed continuous, and discrete variables, some of which may have missing values. Discrete variables are treated with a latent continuous variable approach, and the Dirichlet process is used to construct a mixture model with an unknown number of components. Variable selection is also performed to identify the variables that are most influential for determining cluster membership. The work is motivated by the need to cluster patients thought to potentially have autism spectrum disorder on the basis of many cognitive and/or behavioral test scores. There are a modest number of patients (486) in the data set along with many (55) test score variables (many of which are discrete valued and/or missing). The goal of the work is to (1) cluster these patients into similar groups to help identify those with similar clinical presentation and (2) identify a sparse subset of tests that inform the clusters in order to eliminate unnecessary testing. The proposed approach compares very favorably with other methods via simulation of problems of this type. The results of the autism spectrum disorder analysis suggested 3 clusters to be most likely, while only 4 test scores had high (>0.5) posterior probability of being informative. This will result in much more efficient and informative testing. The need to cluster observations on the basis of many correlated, continuous/discrete variables with missing values is a common problem in the health sciences as well as in many other disciplines.",
keywords = "Dirichlet process, hierarchical Bayesian modeling, missing data, mixed variable types, model-based clustering, variable selection",
author = "Curtis Storlie and Myers, {S. M.} and Katusic, {Slavica K} and Weaver, {A. L.} and Voigt, {R. G.} and Croarkin, {Paul E} and Stoeckel, {R. E.} and Port, {John D}",
year = "2018",
month = "8",
day = "30",
doi = "10.1002/sim.7697",
language = "English (US)",
volume = "37",
pages = "2884--2899",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "19",

}

TY - JOUR

T1 - Clustering and variable selection in the presence of mixed variable types and missing data

AU - Storlie, Curtis

AU - Myers, S. M.

AU - Katusic, Slavica K

AU - Weaver, A. L.

AU - Voigt, R. G.

AU - Croarkin, Paul E

AU - Stoeckel, R. E.

AU - Port, John D

PY - 2018/8/30

Y1 - 2018/8/30

N2 - We consider the problem of model-based clustering in the presence of many correlated, mixed continuous, and discrete variables, some of which may have missing values. Discrete variables are treated with a latent continuous variable approach, and the Dirichlet process is used to construct a mixture model with an unknown number of components. Variable selection is also performed to identify the variables that are most influential for determining cluster membership. The work is motivated by the need to cluster patients thought to potentially have autism spectrum disorder on the basis of many cognitive and/or behavioral test scores. There are a modest number of patients (486) in the data set along with many (55) test score variables (many of which are discrete valued and/or missing). The goal of the work is to (1) cluster these patients into similar groups to help identify those with similar clinical presentation and (2) identify a sparse subset of tests that inform the clusters in order to eliminate unnecessary testing. The proposed approach compares very favorably with other methods via simulation of problems of this type. The results of the autism spectrum disorder analysis suggested 3 clusters to be most likely, while only 4 test scores had high (>0.5) posterior probability of being informative. This will result in much more efficient and informative testing. The need to cluster observations on the basis of many correlated, continuous/discrete variables with missing values is a common problem in the health sciences as well as in many other disciplines.

AB - We consider the problem of model-based clustering in the presence of many correlated, mixed continuous, and discrete variables, some of which may have missing values. Discrete variables are treated with a latent continuous variable approach, and the Dirichlet process is used to construct a mixture model with an unknown number of components. Variable selection is also performed to identify the variables that are most influential for determining cluster membership. The work is motivated by the need to cluster patients thought to potentially have autism spectrum disorder on the basis of many cognitive and/or behavioral test scores. There are a modest number of patients (486) in the data set along with many (55) test score variables (many of which are discrete valued and/or missing). The goal of the work is to (1) cluster these patients into similar groups to help identify those with similar clinical presentation and (2) identify a sparse subset of tests that inform the clusters in order to eliminate unnecessary testing. The proposed approach compares very favorably with other methods via simulation of problems of this type. The results of the autism spectrum disorder analysis suggested 3 clusters to be most likely, while only 4 test scores had high (>0.5) posterior probability of being informative. This will result in much more efficient and informative testing. The need to cluster observations on the basis of many correlated, continuous/discrete variables with missing values is a common problem in the health sciences as well as in many other disciplines.

KW - Dirichlet process

KW - hierarchical Bayesian modeling

KW - missing data

KW - mixed variable types

KW - model-based clustering

KW - variable selection

UR - http://www.scopus.com/inward/record.url?scp=85050138161&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85050138161&partnerID=8YFLogxK

U2 - 10.1002/sim.7697

DO - 10.1002/sim.7697

M3 - Article

AN - SCOPUS:85050138161

VL - 37

SP - 2884

EP - 2899

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 19

ER -