Latent multinomial models for extended batch-mark data.

Area of Science:

• Ecology
• Population Biology
• Statistical Modeling

Background:

• Batch marking is a practical alternative to individual marking in capture-recapture studies due to constraints like cost and difficulty.
• Traditional capture-recapture models struggle with batch-marked data, as observed counts do not represent individual capture histories, leading to computational challenges.
• Existing methods are often computationally infeasible for analyzing data from multiple capture occasions with batch marks.

Purpose of the Study:

• To develop a computationally efficient statistical model for capture-recapture data obtained through batch marking.
• To address the limitations of traditional models when dealing with aggregated count data instead of individual histories.
• To provide a flexible framework applicable to various study designs using batch marks.

Main Methods:

• Proposed a latent multinomial model where observed counts are derived from an underlying multinomial distribution.
• Employed a saddlepoint approximation-based maximum likelihood approach for efficient model fitting.
• Validated the model's performance through simulation studies.

Main Results:

• The latent multinomial model effectively handles count data from batch-marked individuals.
• Simulation studies demonstrated reliable estimation of all model parameters.
• The model was successfully applied to analyze golden mantella population data from Madagascar.

Conclusions:

• The proposed latent multinomial model offers a flexible and efficient solution for capture-recapture studies using batch marks.
• This approach overcomes the computational intractability of traditional methods for such data.
• The model provides accurate parameter estimation and is applicable across diverse ecological study designs.