Bayesian Epidemiologic

Screening Techniques

Research Focus

The following modules contain software, program code, illustrated examples, and associated publications, all of which are provided for the user's convenience. We endeavor here to make available to a larger audience some of the methods we've developed in addressing issues of analyzing data derived from imperfect diagnostic tests. These programs have mainly been written for use with WinBUGS, to facilitate their use by data analysts. The WinBUGS 1.4 software may be downloaded free from the WinBUGS Project, though some S-PLUS and R code is also presented.

The software modules are organized based on the type of data they would be used to analyze. For example, the module "Diagnostic Test Se and Sp Estimation: 2 independent tests, 2 populations, no gold standard (TAGS)" is designed to be used with data that consist of two populations (with different prevalences) that are cross-classified based on the results of two diagnostic tests that are independent conditional on disease status. The parameters about which inferences are made include sensitivity and specificity of each diagnostic test. The module contains an expanded description of the model, instructions for downloading and using "TAGS" software, the journal article in which TAGS was first described, and a worked example of how to use TAGS to analyze a sample dataset.

Unless otherwise stated, the models presented below are Bayesian in nature.

Diagnostic Test Se and Sp Estimation

• 2 independent tests, 2 populations, no gold standard

• 2 independent tests, 2 populations, no gold standard (TAGS) - frequentist approach

• 2 independent tests, 2 populations, no gold standard - spreadsheet workbook to estimate Se and Sp, and for frequentist sample size calculations [Microsoft Excel file: HWsamplesize.xls ]

• 2 dependent tests, 1 population, no gold standard

• 2 dependent tests, 2 populations, no gold standard

• 3 tests (2 dependent and 1 independent), 1 population, no gold standard

• 3 tests (2 dependent and 1 independent), 2 populations, no gold standard

Disease Freedom

• Bayesfreecalc: Posterior probability of herd freedom from disease and sample size calculation

Logistic Regression

• Outcome measured perfectly - ordinary logistic model

• Outcome based on imperfect test results, and measured imperfectly or with “error”

Prevalence Estimation

• 1 test, 1 population - binomial sampling

• 1 test, 1 population - hypergeometric sampling

• 1 test, multiple populations - estimation of herd-level prevalence

• 2 tests, multiple populations - estimation of prevalence distribution

• Posterior distribution of true prevalence given apparent prevalence

• Monte Carlo simulated NPV, PPV, LR+, and LR- given distributions for prevalence, Se, and Sp

• A Bayesian approach to estimate OJD prevalence from pooled faecal samples of variable pool size [WinBUGS file: OJDPrev wo data.odc ]

Prior Elicitation

• Sliders

Receiver Operating Characteristic (ROC) Curve Estimation

• Estimating ROC curves and corresponding AUC’s based on a gold standard

• Estimating ROC curves and corresponding AUC’s in the absence of a gold standard

UC Legal Disclaimer

The Regents of the University of California disclaim all warranties with regard to these software, including all implied warranties of merchantability, non-infringement, and fitness for a particular use. In no event shall the Regents of the University of California be liable for any direct, special, indirect, or consequential damage or any damage whatsoever resulting from the loss of use, data or profits, whether in an action of contract, negligence or other tortious action, arising out of or in connection with the use or performance of these software. Use at your own risk. If you do not agree to this, do not use these software.