\`x^2+y_1+z_12^34\`
Advanced Search
Article Contents
Article Contents

Particle filters for inference of high-dimensional multivariate stochastic volatility models with cross-leverage effects

  • * Corresponding author

    * Corresponding author

AJ is supported by an AcRF tier 2 grant: R-155-000-161-112

Abstract Full Text(HTML) Figure(14) / Table(3) Related Papers Cited by
  • Multivariate stochastic volatility models are a popular and well-known class of models in the analysis of financial time series because of their abilities to capture the important stylized facts of financial returns data. We consider the problems of filtering distribution estimation and also marginal likelihood calculation for multivariate stochastic volatility models with cross-leverage effects in the high dimensional case, that is when the number of financial time series that we analyze simultaneously (denoted by $ d $) is large. The standard particle filter has been widely used in the literature to solve these intractable inference problems. It has excellent performance in low to moderate dimensions, but collapses in the high dimensional case. In this article, two new and advanced particle filters proposed in [4], named the space-time particle filter and the marginal space-time particle filter, are explored for these estimation problems. The better performance in both the accuracy and stability for the two advanced particle filters are shown using simulation and empirical studies in comparison with the standard particle filter. In addition, Bayesian static model parameter estimation problem is considered with the advances in particle Markov chain Monte Carlo methods. The particle marginal Metropolis-Hastings algorithm is applied together with the likelihood estimates from the space-time particle filter to infer the static model parameter successfully when that using the likelihood estimates from the standard particle filter fails.

    Mathematics Subject Classification: Primary: 82C80; Secondary: 60G99, 62F15.

    Citation:

    \begin{equation} \\ \end{equation}
  • 加载中
  • Figure 1.  Plot of Scaled Effective Sample Size (ESS) averaged over 20 runs when standard particle filter is applied to MSV model with dimension 200

    Figure 2.  Plots of Scaled Effective Sample Size (ESS) averaged over 20 runs

    Figure 3.  Plots of mean of estimates for the $ 1^{st} $ component of the mean of the filters across 20 runs

    Figure 4.  Plots of SD of estimates for the $ 1^{st} $ component of the mean of the filters across 20 runs

    Figure 5.  Plots of SD of the estimated log-likelihoods across 20 runs

    Figure 6.  Time Comparison Study Plots of Scaled Effective Sample Size (ESS) averaged over 20 runs

    Figure 7.  Time Comparison Study Plots of Relative SD of the estimated log-likelihoods (w.r.t. the SD of the STPF) across 20 runs

    Figure 8.  Trace plots, histograms and ACF plots of the parameter estimates using PMMH for $ \rho_{ij,\varepsilon\varepsilon} $

    Figure 9.  Trace plots, histograms and ACF plots of the parameter estimates using PMMH for $ \sigma_{i,\eta\eta} $

    Figure 10.  Plot of Scaled Effective Sample Size (ESS)

    Figure 11.  Plots of Scaled Effective Sample Size (ESS) averaged over 20 runs

    Figure 12.  Plots of SD of the estimated log-likelihoods across 20 runs

    Figure 13.  Time Comparison Study Plots of Scaled Effective Sample Size (ESS) averaged over 20 runs

    Figure 14.  Time Comparison Study Plots of Relative SD of the estimated log-likelihoods (w.r.t. the SD of the STPF) across 20 runs

    Table 1.  Number of particles used in each algorithm

    $ d $ Standard PF STPF Marginal STPF
    25 $ N=1000 $ $ N=50 $, $ M_d=20 $ $ N=1 $, $ M_d=1000 $
    50 $ N=1000 $ $ N=50 $, $ M_d=20 $ $ N=1 $, $ M_d=1000 $
    100 $ N=1000 $ $ N=50 $, $ M_d=20 $ N.A.
    200 $ N=1000 $ $ N=50 $, $ M_d=20 $ N.A.
     | Show Table
    DownLoad: CSV

    Table 2.  Computation time (in minutes) per 50 time points for each algorithm

    $ d $ Standard PF STPF Marginal STPF
    25 $ 0.3 $ $ 2 $ $ 46 $
    50 $ 0.6 $ $ 3 $ $ 110 $
    100 $ 2 $ $ 16.7 $ N.A.
    200 $ 5 $ $ 120 $ N.A.
     | Show Table
    DownLoad: CSV

    Table 3.  Number of particles and computation time (in minutes) per 50 time points for each algorithm

    $ d $ Standard PF STPF Computation Time
    100 $ N=25000 $ $ N=50 $, $ M_d=20 $ $ 16.7 $
    200 $ N=40000 $ $ N=50 $, $ M_d=20 $ $ 120 $
     | Show Table
    DownLoad: CSV
  • [1] C. AndrieuA. Doucet and R. Holenstein, Particle Markov chain Monte Carlo methods (with discussion), J. R. Statist. Soc. Ser. B, 72 (2010), 269-342.  doi: 10.1111/j.1467-9868.2009.00736.x.
    [2] M. AsaiM. McAleer and J. Yu, Multivariate stochastic volatility: A review, Econ. Rev., 25 (2006), 145-175.  doi: 10.1080/07474930600713564.
    [3] L. BauwensS. Laurent and J. V. Rombouts, Multivariate GARCH models: A survey, J. Appl. Econ., 21 (2006), 79-109.  doi: 10.1002/jae.842.
    [4] A. BeskosD. CrisanA. JasraK. Kamatani and Y. Zhou, A stable particle filter for a class of high-dimensional state-space models, Adv. Appl. Probab., 49 (2017), 24-48.  doi: 10.1017/apr.2016.77.
    [5] P. Bickel, B. Li and T. Bengtsson, Sharp failure rates for the bootstrap particle filter in high dimensions, In Pushing the Limits of Contemporary Statistics: Contributions in Honor of J. Ghosh, IMS, 3 (2008), 318–329. doi: 10.1214/074921708000000228.
    [6] S. ChibF. Nadari and N. Shephard, Analysis of high-dimensional multivariate stochastic volatility models, J. Econ., 134 (2006), 341-371.  doi: 10.1016/j.jeconom.2005.06.026.
    [7] N. Chopin, Central limit theorem for sequential Monte Carlo methods and its application to Bayesian inference, Ann. Statist., 32 (2004), 2385-2411.  doi: 10.1214/009053604000000698.
    [8] P. Del Moral, Feynman-Kac Formulae: Genealogical and Interacting Particle Systems with Applications, Springer, New York, 2004. doi: 10.1007/978-1-4684-9393-1.
    [9] P. Del MoralA. Doucet and A. Jasra, On adaptive resampling strategies for sequential Monte Carlo methods, Bernoulli, 18 (2012), 252-278.  doi: 10.3150/10-BEJ335.
    [10] A. Doucet, On Sequential Simulation-based Methods for Bayesian Filtering, Technical Report, 1998.
    [11] A. Doucet and A. Johansen, A tutorial on particle filtering and smoothing: Fifteen years later, In Handbook of Nonlinear Filtering (eds. D. Crisan & B. Rozovsky), Oxford University Press, Oxford, (2011), 656–704.
    [12] A. DoucetM. K. PittG. Deligiannidis and R. Kohn, Efficient Implementation of Markov chain Monte Carlo when Using an Unbiased Likelihood Estimator, Biometrika, 102 (2015), 295-313.  doi: 10.1093/biomet/asu075.
    [13] J. Hull and A. White, The pricing of options on assets with stochastic volatilities, J. Finan., 42 (1987), 281-300.  doi: 10.1111/j.1540-6261.1987.tb02568.x.
    [14] T. Ishihara and Y. Omori, Efficient Bayesian estimation of a multivariate stochastic volatility with cross leverage and heavy tailed errors, Comp. Statist. Data Anal., 56 (2012), 3674-3689.  doi: 10.1016/j.csda.2010.07.015.
    [15] A. JasraD. A. StephensA. Doucet and T. Tsagaris, Inference for Lévy driven stochastic volatility models via adaptive sequential Monte Carlo, Scand. J. Statist., 38 (2011), 1-22.  doi: 10.1111/j.1467-9469.2010.00723.x.
    [16] N. KantasA. DoucetS. S. SinghJ. M. Maciejowski and N. Chopin, An overview of sequential Monte Carlo methods for parameter estimation in general state-space sodels, IFAC Proc., 42 (2009), 774-785. 
    [17] N. KantasA. DoucetS. S. SinghJ. M. Maciejowski and N. Chopin, On particle methods for parameter estimation in general state-space models, Statist. Sci., 30 (2015), 328-351.  doi: 10.1214/14-STS511.
    [18] S. KimN. Shephard and S. Chib, Stochastic volatility: Likelihood inference and comparison with ARCH models, Rev. Econ. Stud., 65 (1998), 361-393.  doi: 10.1111/1467-937X.00050.
    [19] G. Kitagawa, Monte Carlo filter and smoother for non-Gaussian nonlinear state-space models, J. Comp. Graph. Stat., 5 (1996), 1-25.  doi: 10.2307/1390750.
    [20] M. Klaas, N. De Freitas and A. Doucet, Towards practical N2 Monte Carlo: The marginal particle filter, Uncert. A. I., (2005), 308–315.
    [21] A. KongJ. S. Liu and W. H. Wong, Sequential imputations and Bayesian missing data problems, J. Amer. Statist. Assoc., 89 (1994), 278-288.  doi: 10.1080/01621459.1994.10476469.
    [22] C. Naesseth, F. Lindten and T. Schön, Nested sequential Monte Carlo methods, ICML, (2015), 1292–1301.
    [23] J. Nakajima, Bayesian analysis of multivariate stochastic volatility with skew return distribution, Econ. Rev., 36 (2017), 546-562.  doi: 10.1080/07474938.2014.977093.
    [24] S. S. Ozturk and J. F. Richard, Stochastic volatility and leverage: Application to a panel of S & P 500 stocks, Finan. Res. Lett., 12 (2015), 67-76. 
    [25] M. K. PittR. Dos Santos SilvaP. Giordani and R. Kohn, On some properties of Markov chain Monte Carlo simulation methods based upon the particle filter, J. Econom., 171 (2012), 134-151.  doi: 10.1016/j.jeconom.2012.06.004.
    [26] M. K. Pitt and N. Shephard, Filtering via simulation: Auxiliary particle filters, J. Amer. Statist. Assoc., 94 (1999), 590-599.  doi: 10.1080/01621459.1999.10474153.
    [27] K. Platanioti, E. McCoy and D. A. Stephens, A Review of Stochastic Volatility Models, Technical Report, 2005.
    [28] C. SnyderT. BengtssonP. Bickel and J. Anderson, Obstacles to high-dimensional particle filtering, Month. Weather Rev., 136 (2008), 4629-4640.  doi: 10.1175/2008MWR2529.1.
    [29] C. VergéC. DuberryP. Del Moral and E. Moulines, On parallel implementation of sequential Monte Carlo methods: The island particle filtering, Stat. Comp., 25 (2015), 243-260.  doi: 10.1007/s11222-013-9429-x.
  • 加载中

Figures(14)

Tables(3)

SHARE

Article Metrics

HTML views(2851) PDF downloads(507) Cited by(0)

Access History

Other Articles By Authors

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return