Estimation in capture-recapture models when covariates are subject to measurement errors.

Area of Science:

• Ecology
• Statistical Modeling
• Biometrics

Background:

• Capture-recapture models are vital for estimating population sizes.
• Covariates measured with errors can introduce significant bias in ecological estimations.
• The naive approach of ignoring measurement errors leads to unacceptable biases and flawed confidence intervals.

Purpose of the Study:

• To address estimation challenges in capture-recapture models with covariate measurement errors.
• To develop robust estimators for regression parameters and population size.
• To compare the performance of a new method against naive and SIMEX approaches.

Main Methods:

• Utilized regression calibration to derive estimators for regression parameters.
• Developed modified estimators for population size accounting for measurement errors.
• Conducted a simulation study to evaluate estimator performance.

Main Results:

• The naive approach, ignoring measurement errors, yields biased estimators and unreliable confidence intervals.
• Regression calibration estimators demonstrated superior performance compared to naive and SIMEX methods.
• Analysis of Prinia flaviventris data highlighted the impact of measurement errors on estimations.

Conclusions:

• Accounting for measurement errors is crucial for accurate capture-recapture estimations.
• The proposed regression calibration method offers a more satisfactory solution for handling covariate errors.
• Findings underscore the importance of addressing measurement error in ecological and biometric studies.