Wesley O. Johnson

Estimation of sensitivity and specificity of diagnostic tests and disease prevalence when the true disease state is unknown

The performance of a new diagnostic test is frequently evaluated by comparison to a perfect reference test (i.e. a gold standard). In many instances, however, a reference test is less than perfect. In this paper, we review methods for estimation of the accuracy of a diagnostic test when an imperfect reference test with known classification errors is available. Furthermore, we focus our presentation on available methods of estimation of test characteristics when the sensitivity and specificity of both tests are unknown. We present some of the available statistical methods for estimation of the accuracy of diagnostic tests when a reference test does not exist (including maximum likelihood estimation and Bayesian inference). We illustrate the application of the described methods using data from an evaluation of a nested polymerase chain reaction and microscopic examination of kidney imprints for detection of Nucleospora salmonis in rainbow trout.

A New Perspective on Priors for Generalized Linear Models

Abstract This article deals with specifications of informative prior distributions for generalized linear models. Our emphasis is on specifying distributions for selected points on the regression surface; the prior distribution on regression coefficients is induced from this specification. We believe that it is inherently easier to think about conditional means of observables given the regression variables than it is to think about model-dependent regression coefficients. Previous use of conditional means priors seems to be restricted to logistic regression with one predictor variable and to normal theory regression. We expand on the idea of conditional means priors and extend these to arbitrary generalized linear models. We also consider data augmentation priors where the prior is of the same form as the likelihood. We show that data augmentation priors are special cases of conditional means priors. With current Monte Carlo methodology, such as importance sampling and Gibbs sampling, our priors result in...

Case-deletion diagnostics for mixed models

Mixed linear models arise in many areas of application. Standard estimation methods for mixed models are sensitive to bizarre observations. Such influential observations can completely distort an analysis and lead to inappropriate actions and conclusions. We develop case-deletion diagnostics for detecting influential observations in mixed linear models. Diagnostics for both fixed effects and variance components are proposed. Computational formulas are given that make the procedures feasible. The methods are illustrated using examples.

Modeling Regression Error With a Mixture of Polya Trees

We model the error distribution in the standard linear model as a mixture of absolutely continuous Polya trees constrained to have median 0. By considering a mixture, we smooth out the partitioning effects of a simple Polya tree and the predictive error density has a derivative everywhere except 0. The error distribution is centered around a standard parametric family of distributions and thus may be viewed as a generalization of standard models in which important, data-driven features, such as skewness and multimodality, are allowed. By marginalizing the Polya tree, exact inference is possible up to Markov chain Monte Carlo error.

Correlation-adjusted estimation of sensitivity and specificity of two diagnostic tests

Models for multiple-test screening data generally require the assumption that the tests are independent conditional on disease state. This assumption may be unreasonable, especially when the biological basis of the tests is the same. We propose a model that allows for correlation between two diagnostic test results. Since models that incorporate test correlation involve more parameters than can be estimated with the available data, posterior inferences will depend more heavily on prior distributions, even with large sample sizes. If we have reasonably accurate information about one of the two screening tests (perhaps the standard currently used test) or the prevalences of the populations tested, accurate inferences about all the parameters, including the test correlation, are possible. We present a model for evaluating dependent diagnostic tests and analyse real and simulated data sets. Our analysis shows that, when the tests are correlated, a model that assumes conditional independence can perform very poorly. We recommend that, if the tests are only moderately accurate and measure the same biological responses, researchers use the dependence model for their analyses. Copyright 2003 Royal Statistical Society.

Comparison of methods for estimation of individual-level prevalence based on pooled samples.

We review frequentist and Bayesian approaches for estimating animal-level disease prevalence using pooled samples obtained by simple random sampling. We determine the preferred approach for different prevalence scenarios and with varying knowledge about sensitivity and specificity values. When sensitivity and specificity are perfect or known, we can choose between the large-sample theory estimates and the one-to-one relationship exact estimates. When sensitivity and specificity are unknown, we must use large-sample theory estimates or Bayesian methodology (which gives exact estimates). However, when the large-sample theory produces a negative lower confidence limit, we must use one of the exact methods. We compare estimates from each approach using culture results from pools of 20 eggs from three flocks on a California ranch that were producing eggs that were contaminated with Salmonella enteritidis phage type 4.

Screening with Cost-Effective Quality Control: Potential Applications to HIV and Drug Testing

Abstract This article investigates the applicability of group testing as a quality control procedure to monitor the sensitivity of screening tests used to check the blood supply for infective agents or to check employees for drug use. The problem is important, as the accuracy of screening tests in the field may deteriorate over time. In the blood screening application, our results demonstrate that group testing the screened negatives provides a procedure with high power to detect a decline of .02 in the sensitivity of the original test when the prevalence in the population is quite low (.0001). Moreover, the procedure is cost-effective in the sense that the expected cost per human immunodeficiency virus (HIV) infection avoided could be less than

Pooled-Sample Testing as a Herd-Screening Tool for Detection of Bovine Viral Diarrhea Virus Persistently Infected Cattle

1 million, in contrast to much higher economic valuations of life that are used in regulatory analyses. The statistical properties of estimates of the prevalence, as well as those for the sensitivity and specificity of the screening test using the extra informati...

Bayesian Binomial Regression: Predicting Survival at a Trauma Center

The study was conducted to develop methodology for least-cost strategies for using polymerase chain reaction (PCR)/probe testing of pooled blood samples to identify animals in a herd persistently infected with bovine viral diarrhea virus (BVDV). Cost was estimated for 5 protocols using Monte Carlo simulations for herd prevalences of BVDV persistent infection (BVDV-PI) ranging from 0.5% to 3%, assuming a cost for a PCR/probe test of

Hierarchical Models for Estimating Herd Prevalence and Test Accuracy in the Absence of a Gold Standard

20. The protocol associated with the least cost per cow involved an initial testing of pools followed by repooling and testing of positive pools. For a herd prevalence of 1%, the least cost per cow was