Using decision trees to understand structure in missing data

被引:36
作者
Tierney, Nicholas J. [1 ,2 ]
Harden, Fiona A. [3 ,4 ]
Harden, Maurice J. [5 ]
Mengersen, Kerrie L. [1 ,2 ]
机构
[1] Queensland Univ Technol, Fac Sci & Engn, Dept Stat Sci, Math Sci, Brisbane, Qld 4001, Australia
[2] ARC Ctr Excellence Math & Stat Frontiers ACEMS, Brisbane, Qld, Australia
[3] Queensland Univ Technol, Fac Hlth, Clin Sci, Brisbane, Qld 4001, Australia
[4] Inst Hlth & Biomed Innovat, Brisbane, Qld, Australia
[5] Hunter Ind Med, Newcastle, NSW, Australia
来源
BMJ OPEN | 2015年 / 5卷 / 06期
基金
澳大利亚研究理事会;
关键词
PATTERN-MIXTURE MODELS; MULTIPLE IMPUTATION; REGRESSION TREES; EXPOSURE; PROGRAM; COHORT;
D O I
10.1136/bmjopen-2014-007450
中图分类号
R5 [内科学];
学科分类号
1002 ; 100201 ;
摘要
Objectives: Demonstrate the application of decision trees-classification and regression trees (CARTs), and their cousins, boosted regression trees (BRTs)-to understand structure in missing data. Setting: Data taken from employees at 3 different industrial sites in Australia. Participants: 7915 observations were included. Materials and methods: The approach was evaluated using an occupational health data set comprising results of questionnaires, medical tests and environmental monitoring. Statistical methods included standard statistical tests and the 'rpart' and 'gbm' packages for CART and BRT analyses, respectively, from the statistical software 'R'. A simulation study was conducted to explore the capability of decision tree models in describing data with missingness artificially introduced. Results: CART and BRT models were effective in highlighting a missingness structure in the data, related to the type of data (medical or environmental), the site in which it was collected, the number of visits, and the presence of extreme values. The simulation study revealed that CART models were able to identify variables and values responsible for inducing missingness. There was greater variation in variable importance for unstructured as compared to structured missingness. Discussion: Both CART and BRT models were effective in describing structural missingness in data. CART models may be preferred over BRT models for exploratory analysis of missing data, and selecting variables important for predicting missingness. BRT models can show how values of other variables influence missingness, which may prove useful for researchers. Conclusions: Researchers are encouraged to use CART and BRT models to explore and understand missing data.
引用
收藏
页数:11
相关论文
共 32 条
[1]  
[Anonymous], 2013, INTRO STAT LEARNING
[2]  
[Anonymous], VIM VISUALIZATION IM
[3]  
[Anonymous], 1997, 61 SECT BIOST MAYO C
[4]  
[Anonymous], MISSINGDATAGUI
[5]  
[Anonymous], 1996, Journal of Computational and Graphical Statistics, DOI [DOI 10.2307/1390776, 10.1080/106 18600.1996.10474700., DOI 10.1080/10618600.1996.10474700]
[6]   SmcHD1, containing a structural-maintenance-of-chromosomes hinge domain, has a critical role in X inactivation [J].
Blewitt, Marnie E. ;
Gendrel, Anne-Valerie ;
Pang, Zhenyi ;
Sparrow, Duncan B. ;
Whitelaw, Nadia ;
Craig, Jeffrey M. ;
Apedaile, Anwyn ;
Hilton, Douglas J. ;
Dunwoodie, Sally L. ;
Brockdorff, Neil ;
Kay, Graham F. ;
Whitelaw, Emma .
NATURE GENETICS, 2008, 40 (05) :663-669
[7]   Statistical modeling: The two cultures [J].
Breiman, L .
STATISTICAL SCIENCE, 2001, 16 (03) :199-215
[8]   A working guide to boosted regression trees [J].
Elith, J. ;
Leathwick, J. R. ;
Hastie, T. .
JOURNAL OF ANIMAL ECOLOGY, 2008, 77 (04) :802-813
[9]   Multiple additive regression trees with application in epidemiology [J].
Friedman, JH ;
Meulman, JJ .
STATISTICS IN MEDICINE, 2003, 22 (09) :1365-1381
[10]   Missing Data Analysis: Making It Work in the Real World [J].
Graham, John W. .
ANNUAL REVIEW OF PSYCHOLOGY, 2009, 60 :549-576