Maximizing the usefulness of statistical classifiers for two populations with illustrative applications

Abstract

The usefulness of two-class statistical classifiers is limited when one or both of the conditional misclassification rates is unacceptably high. Incorporating a neutral zone region into the classifier provides a mechanism to refer ambiguous cases to follow-up where additional information might be obtained to clarify the classification decision. Through the use of the neutral zone region, the conditional misclassification rates can be controlled and the classifier becomes useful. Three real-life examples, including applications to prostate cancer and kidney dysfunction following heart surgery, are used to illustrate how neutral zone regions can extract utility from disappointing classifiers that might otherwise be abandoned.

Keywords

Statistical classification logistic regression neutral zones receiving operating characteristic curve prostate cancer

Get full access to this article

View all access options for this article.

References

James

Witten

Hastie

et al.

An introduction to statistical learning, New York: Springer, 2013.

Fawcett

. An introduction to ROC analysis. Pattern Recogn Lett 2006; 27: 861–874.

Baker

. The central role of receiver operating characteristic (ROC) curves in evaluating tests for the early detection of cancer. J Natl Cancer Inst 2003; 95: 511–515.

Youden

. Index for rating diagnostic tests. Cancer 1950; 3: 32–35.

Ripley

. Pattern recognition and neutral networks, UK: Cambridge University Press, 1996.

Rao

. The problem of classification and distance between two populations. Nature 1947; 18: 30–31.

Habbema

JDF

Hermans

Van Der Brugt

. Cases of doubt in allocation problems. Biometrika 1974; 61: 313–324.

Jeske

Liu

Bent

et al.

Classification rules that include neutral zones and their application to microbial community profiling. Commun Stat – Theory Meth 2007; 36: 1965–1980.

Jeske

Ruegger

et al.

A three-class neutral classifier using a decision-theoretic approach with application to DNA array analyses. J Agri Environ Stat 2010; 15: 474–490.

10.

Benecke

Jeske

Reugger

et al.

Bayes neutral zone classifiers with applications to nonparametric unsupervised settings. J Agri Biol Environ Stat 2013; 18: 39–52.

11.

Chow

. An optimum character recognition system using decision functions. IRE Trans Electron Comput 1957; 4: 247–254.

12.

Anderson

. Constrained discrimination between k populations. J Roy Stat Soc Ser B 1969; 31: 123–139.

13.

Patterson

. A three-population constrained discrimination procedure. Commun Stat – Theory Meth 2016; 45: 4771–4787.

14.

Linehan

Wilson

Kawachi

et al.

Pre-surgical prediction of early biochemical recurrence with expressed prostatic secretion (EPS) biomarkers. J Clin Oncol 2016; 34. abstract #16593.

15.

Chandra

Roy

. Some results on reversed hazard rate. Prob Inform Sci 2001; 15: 95–102.

16.

Bishoff

Freedland

Gerber

et al.

Prognostic utility of the cell cycle progression score generated from biopsy in men treated with prostatectomy. J Urol 2014; 192: 409–414.

17.

Zhang

Jeske

et al.

A sequential logistic regression classifier based on mixed effects with applications to longitudinal data. Comput Stat Data Anal 2016; 94: 238–249.

18.

Bagnoli

Bergstrom

. Log-concave probability and its applications. Econom Theory 2005; 26: 445–469.

19.

Al-Zahrani

Stoyanov

. On some properties of life distributions with increasing elasticity and log-concavity. Appl Math Sci 2008; 2: 2349–2361.

20.

Gupta

Nanda

. Some results on reversed hazard rate ordering. Commun Stat – Theory Meth 2001; 30: 2447–2457.

21.

Harville

. Matrix algebra from a statistician’s perspective, New York: Springer, 1997.