Computational Efficiency and Precision for Replicated-Count and Batch-Marked Hidden Population Models.

Area of Science:

• Ecology
• Computational Biology
• Statistical Modeling

Background:

• Open-population N-mixture models, hidden integer-valued autoregressive models, and hidden Markov models face computational challenges.
• These challenges include long computation times and difficulties with model likelihood functions involving many hidden states.

Purpose of the Study:

• To present computationally efficient and numerically stable methods for analyzing ecological models.
• To improve the precision and speed of ecological population size estimation.

Main Methods:

• Utilizing fast Fourier transforms to accelerate computations.
• Implementing improved numerical stability techniques for model likelihood calculations.
• Applying these methods to open-population N-mixture models.

Main Results:

• Achieved significant speed improvements (up to 100x) compared to state-of-the-art ecological software for N-mixture models.
• Demonstrated successful estimation of a large elk population size where previous software failed due to numerical issues.
• Showcased enhanced precision for likelihood functions involving sums of logarithms and improved computational efficiency for models with convolutions.

Conclusions:

• The developed methods offer substantial improvements in computational efficiency and numerical stability for various ecological models.
• These advancements enable more precise and faster estimation of population sizes, particularly for large datasets or complex models.
• The techniques are broadly applicable to ecological modeling, addressing common computational bottlenecks.