SEQUENTIAL PARTICLE METHODS FOR BAYESIAN EVALUATIONS OF ABUNDANCE AND CATCH AT AGE

A cohort population dynamics may be represented as a hidden Bayesian model with abundances as hidden states and catches as observations. Using these models, one can evaluate posterior densities and calculate such point-wise characteristics as means, medians, variances and so on. With rare exceptions (as linear Gaussian models), the recurrence equations met by the posterior densities have no analytic solutions. We describe several particle (Monte Carlo) methods that may be used for the density approximations and evaluations of their statistical quantities for nonlinear non-Gaussian models as well. The Fishmetica package was extended with functions for generating samples and masses for time series filtering, prediction, and smoothing. Evaluations in Julia were made for a test dataset with assumed Gaussian distributions of residuals and different numbers of particles. Numerical results were compared with known analytic ones.

оценка промысловых запасов, пространство состояний, скрытые марковские сети, метод частиц, последовательное оценивание с учетом значимости, фильтрация, предсказание, сглаживание, state-space assessment models, dynamic Bayesian networks, filtering, smoothing, prediction

1. Шевченко И. И. Моделирование промысловых запасов при известных оценках возрастной структуры популяций и уловов // Вопр. рыболовства. 2017. Т. 18. № 4. С. 507- 519.

2. Шевченко И. И. Моделирование промысловых запасов при известных оценках возрастной структуры популяций и уловов. II // Вопр. рыболовства. 2019. Т. 20. № 2. С. 152-163.

3. Aldrin M., Aanes S., Subbey S. Comments on incongruous formulations in the sam (state-space assessment model) model and con sequences for fish stock assessment // Fisheries Research, 2019. № 210. P. 224-227.

4. Crisan D., Doucet A. A survey of convergence results on particle filtering methods for practitioners / / IEEE Transactions on signal processing. 2002. V. 50. № 3. P. 736-746.

5. Doucet A., Godsill S., Andrieu C. On sequential monte carlo sampling methods for bayesian filtering // Statistics and computing. 2000. V. 10. № 3. P. 197-208.

6. Godsill S. J, Doucet A, West M. Monte carlo smoothing for nonlinear time series // J. American statistical association, 2004. V. 99. № 465. P. 156-168.

7. Gordon N.,Salmond D. J.,Smith A. F. M. Novel approach to nonlinear/ non-gaussian bayesian state estimation // IEE Proceedings F (Radar and Signal Processing). 1993. V. 140. P. 107-113.

8. Hurzeler M., Kunsch H. R. Monte carlo approximations for general state-space models // J. Computational and Graphical Statistics. 1998. V. 7. № 2. P. 175-193.

9. Kitagawa G. Monte carlo filter and smoother for non-gaussian nonlinear state space models // J. computational and graphical statistics. 1996. V. 5. № 1. P. 1-25.

10. Kotz S., Kozubowski T. J., Podgorski K. The Laplace Distribution and Generalizations: A Revisit with Applications to Communications, Economics, Engineering, and Finance. Basel: Birkhauser. 2001. 349 p.

11. Nielsen A, Berg C.W. Response to: Comments on incongruous formulations in the sam (state-space assessment model) model and consequences for fish stock assessment // Fisheries Research. 2019. № 210. P. 228229.

12. Russell S. J., Norvig P. Artificial Intelligence: A Modern Approach. Upper Saddle River: Prentice Hall, 2010. 1152 p.

13. Sarkka S. Bayesian filtering and smoothing. Cambridge: Cambridge University Press, 2013. 254 p.

14. Schnute J. T. A general framework for developing sequential fisheries models // Canadian J. Fisheries and Aquatic Sciences. 1994. V. 51. № 8. P. 1676-1688.

15. Smith A. F. M, Gelfand A. E. Bayesian statistics without tears: a sampling-resampling perspective // The American Statistician. 1992. V. 46. № 2. P. 84-88.

16. Valpine P. de, Hilborn R. State-space likelihoods for nonlinear fisheries time-series // Canadian J. Fisheries and Aquatic Sciences. 2005. V. 62. № 9. P. 1937-1952.