Stable distributions are well-known for their desirable properties and can effectively fit data with heavy tail. However, due to the lack of an explicit probability density function and finite second moments in most cases, traditional parametric inference methods are no longer applicable. Bayesian Synthetic Likelihood is a likelihood-free Bayesian inference method based on model simulations, which effectively addresses parameter inference problems when the probability density function is not explicitly available. Semi-parametric Bayesian Synthetic Likelihood relaxes the normality assumption by incorporating semi-parametric estimation methods, but it performs poorly when applied to data with heavy tail and excess kurtosis. To improve this, we introduced adaptive Monte Carlo algorithm to enhance the convergence speed, and transformed kernel density estimation to increase the estimation accuracy. Numerical experiments and empirical analysis on stable distributions validated the superiority of the proposed improvements.
References
[1]
Sisson, S.A., Fan, Y. and Tanaka, M.M. (2007) Sequential Monte Carlo without Likelihoods. Proceedings of the National Academy of Sciences, 104, 1760-1765. https://doi.org/10.1073/pnas.0607208104
[2]
Price, L.F., Drovandi, C.C., Lee, A. and Nott, D.J. (2017) Bayesian Synthetic Likelihood. Journal of Computational and Graphical Statistics, 27, 1-11. https://doi.org/10.1080/10618600.2017.1302882
[3]
Picchini, U. and Forman, J.L. (2019) Bayesian Inference for Stochastic Differential Equation Mixed Effects Models of a Tumour Xenography Study. Journal of the Royal Statistical Society Series C: Applied Statistics, 68, 887-913. https://doi.org/10.1111/rssc.12347
[4]
Maraia, R., Springer, S., Härkönen, T., Simon, M. and Haario, H. (2023) Bayesian Synthetic Likelihood for Stochastic Models with Applications in Mathematical Finance. Frontiers in Applied Mathematics and Statistics, 9, Article 1187878. https://doi.org/10.3389/fams.2023.1187878
[5]
Moriña, D., Fernández-Fontelo, A., Cabaña, A., Arratia, A. and Puig, P. (2023) Estimated Covid-19 Burden in Spain: ARCH Underreported Non-Stationary Time Series. BMC Medical Research Methodology, 23, Article No. 75. https://doi.org/10.1186/s12874-023-01894-9
[6]
Fasiolo, M., Wood, S.N., Hartig, F. and Bravington, M.V. (2018) An Extended Empirical Saddlepoint Approximation for Intractable Likelihoods. Electronic Journal of Statistics, 12, 1544-1578. https://doi.org/10.1214/18-ejs1433
[7]
An, Z., Nott, D.J. and Drovandi, C. (2019) Robust Bayesian Synthetic Likelihood via a Semi-Parametric Approach. Statistics and Computing, 30, 543-557. https://doi.org/10.1007/s11222-019-09904-x
[8]
An, Z., South, L.F. and Drovandi, C. (2022) BSL: An R Package for Efficient Parameter Estimation for Simulation-Based Models via Bayesian Synthetic Likelihood. Journal of Statistical Software, 101, 1-33. https://doi.org/10.18637/jss.v101.i11
[9]
Picchini, U., Simola, U. and Corander, J. (2023) Sequentially Guided MCMC Proposals for Synthetic Likelihoods and Correlated Synthetic Likelihoods. Bayesian Analysis, 18, 1099-1129. https://doi.org/10.1214/22-ba1305
[10]
Silverman, B.W. (1986) Density Estimation for Statistics and Data Analysis. CRC Press.
[11]
Boudt, K., Cornelissen, J. and Croux, C. (2011) The Gaussian Rank Correlation Estimator: Robustness Properties. Statistics and Computing, 22, 471-483. https://doi.org/10.1007/s11222-011-9237-0
[12]
Hastings, W.K. (1970) Monte Carlo Sampling Methods Using Markov Chains and Their Applications. Biometrika, 57, 97-109. https://doi.org/10.1093/biomet/57.1.97
[13]
Haario, H., Saksman, E. and Tamminen, J. (2001) An Adaptive Metropolis Algorithm. Bernoulli, 7, 223-242. https://doi.org/10.2307/3318737
[14]
Gelman, A., Roberts, G.O. and Gilks, W.R. (1996) Efficient Metropolis Jumping Rules. In: Bernardo, J.M., et al., Eds., Bayesian Statistics 5, Oxford University Press, 599-608. https://doi.org/10.1093/oso/9780198523567.003.0038
[15]
Haario, H., Saksman, E. and Tamminen, J. (2001) An adaptive Metropolis Algorithm. Bernoulli, 7, 223-242. https://doi.org/10.2307/3318737
[16]
Gelman, A., Carlin, J.B., Stern, H.S., et al. (2004) Bayesian Data Analysis. 2nd Edition, CRC Press.
[17]
Wand, M.P., Marron, J.S. and Ruppert, D. (1991) Transformations in Density Estimation. Journal of the American Statistical Association, 86, 343-353. https://doi.org/10.1080/01621459.1991.10475041
[18]
Park, B.U., Chung, S.S. and Seog, K.H. (1992) An Empirical Investigation of the Shifted Power Transformation Method in Density Estimation. Computational Statistics & Data Analysis, 14, 183-191. https://doi.org/10.1016/0167-9473(92)90172-c
[19]
Bolancé, C., Guillen, M. and Nielsen, J.P. (2003) Kernel Density Estimation of Actuarial Loss Functions. Insurance: Mathematics and Economics, 32, 19-36. https://doi.org/10.1016/s0167-6687(02)00191-9
[20]
Clements, A., Hurn, S. and Lindsay, K. (2003) Mobius-Like Mappings and Their Use in Kernel Density Estimation. Journal of the American Statistical Association, 98, 993-1000. https://doi.org/10.1198/016214503000000945
[21]
Buch-larsen, T., Nielsen, J.P., Guillén, M. and Bolancé, C. (2005) Kernel Density Estimation for Heavy-Tailed Distributions Using the Champernowne Transformation. Statistics, 39, 503-516. https://doi.org/10.1080/02331880500439782
[22]
Bolancé, C. (2010) Optimal Inverse β(3, 3) Transformation in Kernel Density Estimation. Sort-Statistics and Operations Research Transactions, 34, 223-238.
[23]
Terrell, G.R. and Scott, D.W. (1985) Oversmoothed Nonparametric Density Estimates. Journal of the American Statistical Association, 80, 209-214. https://doi.org/10.1080/01621459.1985.10477163