Closed-Population Size Estimation Approaches for Capture-recapture Models with Missing Covariate Data

Abstract

We consider the problem of closed-population size estimation for capture recapture models when the covariates may be missing at random (MAR). Assuming a logistic regression model for the capture probabilities on each capture occasion, we provide new approaches to account for the missing covariate data. More specifically, we develop new semiparametric estimators for the regression parameters based on multiple imputation and inverse probability weighting techniques. The proposed methods require no distributional assumption about the missing covariates. Large sample properties of proposed estimators are studied under certain regularity conditions. The finite-sample size performance of our estimators are investigated through a simulation study. We apply the proposed methods to real-world capture-recapture data.Key words: Capture recapture models; Horvitz-Thompson; Multiple imputation; Inverse probability weighting; Logistic regression; Missing values; Population size; Regression calibration.