Book Materials

Datasets and Stata Programs used in "The Statistical Evaluation of Medical Tests for Classification and Prediction"

The Statistical Evaluation of Medical Tests for Classification and Prediction book describes statistical concepts and techniques for evaluating medical diagnostic tests and biomarkers for detecting disease. More generally, the techniques pertain to the statistical classification problem for predicting a dichotomous outcome. Measures for quantifying test accuracy are described including sensitivity, specificity, predictive values, diagnostic likelihood ratios and the Receiver Operating Characteristic Curve that is commonly used for continuous and ordinal valued tests. Statistical procedures are presented for estimating and comparing them. Regression frameworks for assessing factors that influence test accuracy and for comparing tests while adjusting for such factors are presented.

This book presents many worked examples of real data and should be of interest to practicing statisticians or quantitative researchers involved in the development of tests for classification or prediction in medicine.

Table of Contents

  1. Introduction
  2. Measures of Accuracy for Binary Tests
  3. Comparing Binary Tests and Regression Analysis
  4. The Receiver Operating Characteristic Curve
  5. Estimating the ROC Curve
  6. Covariate Effects on Continuous and Ordinal Tests
  7. Incomplete Data and Imperfect Reference Tests
  8. Study Design and Hypothesis Testing
  9. More Topics and Conclusions
    References/Bibliography
    Index

Datasets

Study Reference Stata File ASCII File
CASS Leisenring et al. (2000)
Weiner et al. (1979)
est1.dta est1.csv
est1_desc.txt
Pancreatic Ca biomarkers Wieand et al. (1989) wiedat2b.dta wiedat2b.csv
wiedat2b_desc.txt
Ultrasound for hepatic mets Tosteson and Begg. (1988) tostbegg2.dta tostbegg2.csv
tostbegg2_desc.txt
CARET PSA Etzioni et al. (1999) psa2b.dta psa2b.csv
psa2b_desc.txt
Gene expression array Pepe et al. (2003) orchratio2.dta orchratio2.csv
orchratio2_desc.txt
Norton neonatal audiology Norton et al. (2000) nnhs.dta nnhs.csv
nnhs_desc.txt
Leisenring neonatal audiology Leisenring et al. (1997) lplaudio_b.dta lplaudio_b.csv
lplaudio_b_desc.txt
Prostate Ca - St. Louis Smith et al. (1997) psa_dre_v2.dta psa_dre_v2.csv
psa_dre_v2_desc.txt
Stover audiology Stover et al. (1996) dp2.dta dp2.csv
dp2_desc.txt
Scintigraphy study Muller et al. (1989) mlt1.dta mlt1.csv
mlt1_desc.txt
59 Pap screen studies Fahey et al. (1995) fim.dta fim.csv
fim_desc.txt
Prenatal screen data (hypothetical)   hpns.dta hpns.csv
hpns_desc.txt

Stata format data files can be read with versions 8 and above.
Comma-separated ASCII (csv) files include variable names on the first row.


Dataset References

Etzioni R, Pepe M, Longton G, Hu C, Goodman G (1999). Incorporating the time dimension in receiver operating characteristic curves: A case study of prostate cancer. Medical Decision Making 19:242-51.

Fahey MT, Irwig LM, Macaskill P (1995). Meta-analysis of Pap test accuracy. American Journal of Epidemiology 141:680-9.

Leisenring W, Alonzo T, Pepe MS (2000). Comparisons of predictive values of binary medical diagnostic tests for paired designs. Biometrics 56:345-51.

Leisenring W, Pepe MS, Longton G (1997). A marginal regression modelling framework for evaluating medical diagnostic tests. Statistics in Medicine 16:1263-81.

Muller C, Wasserman HJ, Erlank P, Klopper JF, Morkel HR, Ellmann A (1989). Optimisation of density and contrast yielded by multiformat photographic images used for scintigraphy. Physics in Medicine and Biology 34:473-81.

Norton SJ, Gorga MP, Widen JE, Folsom RC, Sininger Y, Cone-Wesson B, Vohr BR, Mascher K, Fletcher K. (2000). Identification of neonatal hearing impairment: Evaluation of transient evoked ototacoustic emission, distortion product otoacoustic emission, and auditory brain stem response test performance. Ear and Hearing 21:508-28.

Pepe MS, Longton G, Anderson G, Schummer M (2003). Selecting differentially expressed genes from microarray experiments. Biometrics (in press) .

Smith DS, Bullock AD, Catalona WJ (1997). Racial differences in operating characteristics of prostate cancer screening tests. The Journal of Urology 158:1861-66.

Stover L, Gorga MP, Neely T (1996). Torwards optimizing the clinical utility of distortion product otoacoustic emission measurements. Journal of the Acoustical Society of America 100:956-967.

Tosteson AN, Begg CB (1988). A general regression methodology for ROC curve estimation. Medical Decision Making 8:204-15.

Weiner DA, Ryan TJ, McCabe CH, Kennedy JW, Schloss M, Tristani F, Chaitman BR, Fisher LD (1979). Exercise stress testing. Correlations among history of angina, ST-segment response and prevalence of coronary-artery disease in the Coronary Artery Aurgery Study (CASS). New England Journal of Medicine 301(5):230-5.

Wieand S, Gail MH, James BR, James KL (1989). A family of nonparametric statistics for comparing diagnostic markers with paired or unpaired data. Biometrika 76:585-92.


Programs

Downloadable Stata programs and help files

Stata version 7 or higher required for most programs; version 8 or 9 required for some as updates and additions become available.

  • emroc.ado, emroc.hlp - Plot the empirical ROC curve and optionally return plot coordinates. Calculate a nonparametric estimate of the area under the ROC curve (AUC) or partial AUC.
  • dfroc.ado, dfroc.hlp - Calculate the distribution-free estimator of the ROC curve within a GLM binary regression framework. Obtain bootstrap standard error estimates for the binormal ROC parameters and correponding AUC.
  • aucbs.ado, aucbs.hlp - Calculate a nonparametric estimate of the area under the ROC curve (AUC) and bootstrapped standard error estimates. Optionally calculate the partial AUC or empirical ROC(t) for specified t and corresponding se estimates. With data for two test measures, difference statistics for the AUC, pAUC, and ROC(t)) are calculated.
  • rocsize.ado, rocsize.hlp - Determine power for a one-sample screening study; continuous data.
  • aucsize.ado, aucsize.hlp - Determine power for a one-sample screening study based on ROC area under the curve (AUC) improvement
  • scrsize.ado, scrsize.hlp - Determine power for a one-sample screening study; binary test outcome data.
  • binscrn1.ado, binscrn1.hlp - Calculates summary screening measures for a test with binary outcome.
  • binscrn2.ado, binscrn2.hlp - Comparison of 2 binary screening tests; for unpaired data.
  • binscrn3.ado, binscrn3.hlp - Comparison of 2 binary screening tests; for paired data.
  • lrreg.ado, lrreg_ll.ado, lrreg.hlp - Diagnostic Likelihood Ratio (DLR) regression.

Utility programs used by text example do-files

  • binormroc.ado, binormroc.hlp - Plots the binormal ROC for specified normal case and control distributions of a test measure.
  • bvnellip.ado, bvnellip.hlp - Calculates a confidence ellipse for the joint distribution of 2 parameters. The parameters are assumed to have a bivariate normal distribution.
  • semt_profile.ado - Specifies data directory path and log file target path for text example do-files.

Examples

Stata do-files for selected text examples and corresponding figures

Chapter 2

Chapter 3

Chapter 4

Chapter 5

Chapter 6

Chapter 7

Chapter 9

Book Errata