Measurement Errors in Capture-Recapture Models with Heterogeneity

Abstract

Measurement error is an important and common problem in epidemiological, medical, and many other disciplines. It is well known that measurement error may cause bias in regression analysis and subsequently lead to invalid statistical inference. There is a growing body of literature investigating measurement error problems. However, most studies of this area to date have their focus primarily on the analysis of regression coefficients in generalized linear models or in failure time regression models, and measurement error in the context of capture-recapture studies is almost never discussed. Here we consider estimation problems in capture-recapture models when the covariates or the auxiliary variables are measured with errors. The naive approach, which ignores measurement errors, is found to be unacceptable in the estimation of both regression parameters and population size. To account for measurement errors, we propose a new conditional score estimation technique to adjust for measurement error in covariate variables for capture-recapture models. Intensive simulation studies are conducted to evaluate the performance of the proposed estimator along with other existing methods. An example involving the bird species Prinia flaviventris is used to illustrate this method. Under most simulation scenarios, the new method is preferable since it has smaller biases and better coverage probabilities. This is a joint work with Steve Y. H. Huang at Tamkang University and C. Y. Wang at Fred Hutchinson Cancer Research Center.