Annales de l'Institut Henri Poincaré, Probabilités et Statistiques

Estimation of the transition density of a Markov chain

Mathieu Sart

Full-text: Open access

Abstract

We present two data-driven procedures to estimate the transition density of an homogeneous Markov chain. The first yields a piecewise constant estimator on a suitable random partition. By using an Hellinger-type loss, we establish non-asymptotic risk bounds for our estimator when the square root of the transition density belongs to possibly inhomogeneous Besov spaces with possibly small regularity index. Some simulations are also provided. The second procedure is of theoretical interest and leads to a general model selection theorem from which we derive rates of convergence over a very wide range of possibly inhomogeneous and anisotropic Besov spaces. We also investigate the rates that can be achieved under structural assumptions on the transition density.

Résumé

Nous présentons deux procédures pour estimer la densité de transition d’une chaîne de Markov homogène. Dans la première procédure, nous construisons un estimateur constant par morceaux sur une partition aléatoire bien choisie. Nous établissons des bornes de risque non-asymptotiques pour une perte de type Hellinger lorsque la racine carrée de la densité de transition appartient à un espace de Besov inhomogène dont l’indice de régularité peut être petit. Nous illustrons ces résultats par des simulations numériques. La deuxième procédure est d’intérêt théorique. Elle permet d’obtenir un théorème de sélection de modèle à partir duquel nous déduisons des vitesses de convergence sur des espaces de Besov inhomogènes anisotropes. Nous étudions finalement les vitesses qui peuvent être atteintes sous des hypothèses structurelles sur la densité de transition.

Article information

Source
Ann. Inst. H. Poincaré Probab. Statist., Volume 50, Number 3 (2014), 1028-1068.

Dates
First available in Project Euclid: 20 June 2014

Permanent link to this document
https://projecteuclid.org/euclid.aihp/1403277007

Digital Object Identifier
doi:10.1214/13-AIHP551

Mathematical Reviews number (MathSciNet)
MR3224298

Zentralblatt MATH identifier
1298.62144

Subjects
Primary: 62M05: Markov processes: estimation 62G05: Estimation

Keywords
Adaptive estimation Markov chain Model selection Robust tests Transition density

Citation

Sart, Mathieu. Estimation of the transition density of a Markov chain. Ann. Inst. H. Poincaré Probab. Statist. 50 (2014), no. 3, 1028--1068. doi:10.1214/13-AIHP551. https://projecteuclid.org/euclid.aihp/1403277007


Export citation

References

  • [1] N. Akakpo. Estimation adaptative par sélection de partitions en rectangles dyadiques. Ph.D. thesis, Univ. Paris Sud, 2009.
  • [2] N. Akakpo. Adaptation to anisotropy and inhomogeneity via dyadic piecewise polynomial selection. Math. Methods Statist. 21 (2012) 1–28.
  • [3] N. Akakpo and C. Lacour. Inhomogeneous and anisotropic conditional density estimation from dependent data. Electron. J. Statist. 5 (2011) 1618–1653.
  • [4] K. B. Athreya and G. S. Atuncar. Kernel estimation for real-valued Markov chains. Sankhyā 60 (1998) 1–17.
  • [5] Y. Baraud. Estimator selection with respect to Hellinger-type risks. Probab. Theory Related Fields 151 (2011) 353–401.
  • [6] Y. Baraud and L. Birgé. Estimating the intensity of a random measure by histogram type estimators. Probab. Theory Related Fields 143 (2009) 239–284.
  • [7] Y. Baraud and L. Birgé. Estimating composite functions by model selection. Ann. Inst. Henri Poincaré Probab. Stat. 50 (2014) 285–314.
  • [8] A. K. Basu and D. K. Sahoo. On Berry–Esseen theorem for nonparametric density estimation in Markov sequences. Bull. Inform. Cybernet. 30 (1998) 25–39.
  • [9] L. Birgé. Approximation dans les espaces métriques et théorie de l’estimation. Probab. Theory Related Fields 65 (1983) 181–237.
  • [10] L Birgé. Stabilité et instabilité du risque minimax pour des variables indépendantes équidistribuées. Ann. Inst. Henri Poincaré Probab. Stat. 20 (1984) 201–223.
  • [11] L. Birgé. Sur un théorème de minimax et son application aux tests. Probab. Math. Statist. 2 (1984) 259–282.
  • [12] L. Birgé. Model selection via testing: An alternative to (penalized) maximum likelihood estimators. Ann. Inst. Henri Poincaré Probab. Stat. 42 (2006) 273–325.
  • [13] L. Birgé. Model selection for Poisson processes. In Asymptotics: Particles, Processes and Inverse Problems 32–64. IMS Lecture Notes Monogr. Ser. 55. IMS, Beachwood, OH, 2007.
  • [14] L. Birgé. Model selection for density estimation with $\mathbb{L}_{2}$-loss. Probab. Theory Related Fields 158 (2014) 533–574.
  • [15] L. Birgé. Robust tests for model selection. In From Probability to Statistics and Back: High-Dimensional Models and Processes. A Festschrift in Honor of Jon Wellner 47–64. IMS Collections 9. IMS, Beachwood, OH, 2012.
  • [16] G. Blanchard, C. Schäfer and Y. Rozenholc. Oracle Bounds and Exact Algorithm for Dyadic Classification Trees. Lecture Notes in Comput. Sci. 3120. Springer, Berlin, 2004.
  • [17] R. C. Bradley. Basic properties of strong mixing conditions. A survey and some open questions. Probab. Surv. 2 (2005) 107–144.
  • [18] S. Clémencon. Adaptive estimation of the transition density of a regular Markov chain. Math. Methods Statist. 9 (2000) 323–357.
  • [19] F. Comte and Y. Rozenholc. Adaptive estimation of mean and volatility functions in (auto-)regressive models. Stochastic Process. Appl. 97 (2002) 111–145.
  • [20] W. Dahmen, R. DeVore and K. Scherer. Multi-dimensional spline approximation. SIAM J. Numer. Anal. 17 (1980) 380–402.
  • [21] R. DeVore and X. Yu. Degree of adaptive approximation. Math. Comput. 55 (1990) 625–635.
  • [22] C. C. Y. Dorea. Strong consistency of kernel estimators for Markov transition densities. Bull. Braz. Math. Soc. (N.S.) 33 (2002) 409–418.
  • [23] P. Doukhan. Mixing: Properties and Examples. Lecture Notes in Statistics 85. Springer, New York, 1994.
  • [24] P. Doukhan and M. Ghindès. Estimation de la transition de probabilité d’une chaîne de Markov Doëblin-récurrente. Étude du cas du processus autorégressif général d’ordre $1$. Stochastic Process. Appl. 15 (1983) 271–293.
  • [25] R. Hochmuth. Wavelet characterizations for anisotropic Besov spaces. Appl. Comput. Harmon. Anal. 12 (2002) 179–208.
  • [26] A. Juditsky, O. Lepski and A. Tsybakov. Nonparametric estimation of composite functions. Ann. Statist. 37 (2009) 1360–1404.
  • [27] C. Lacour. Adaptive estimation of the transition density of a Markov chain. Ann. Inst. Henri Poincaré Probab. Statist. 43 (2007) 571–597.
  • [28] C. Lacour. Nonparametric estimation of the stationary density and the transition density of a Markov chain. Stochastic Process. Appl. 118 (2008) 232–260.
  • [29] C. Lacour. Erratum to “Nonparametric estimation of the stationary density and the transition density of a Markov chain” [Stochastic Process. Appl. 118 (2008) 232–260] []. Stochastic Process. Appl. 122 (2012) 2480–2485.
  • [30] L. Le Cam. Convergence of estimates under dimensionality restrictions. Ann. Statist. 1 (1973) 38–53.
  • [31] L. Le Cam. On local and global properties in the theory of asymptotic normality of experiments. In Stochastic Processes and Related Topics (Proc. Summer Res. Inst. Statist. Inference for Stochastic Processes, Indiana Univ., Bloomington, Ind., 1974, Vol. 1; dedicated to Jerzy Neyman) 13–54. Academic Press, New York, 1975.
  • [32] P. Massart. Concentration Inequalities and Model Selection. Lecture Notes in Mathematics 1896. Springer, Berlin, 2003.
  • [33] G. G. Roussas. Nonparametric estimation in Markov processes. Ann. Inst. Statist. Math. 21 (1969) 73–87.
  • [34] G. G. Roussas. Estimation of Transition Distribution Function and Its Quantiles in Markov Processes: Strong Consistency and Asymptotic Normality. NATO Adv. Sci. Inst. Ser. C Math. Phys. Sci. 335. Kluwer Acad. Publ., Dordrecht, 1991.
  • [35] M. Sart. Model selection for poisson processes with covariates. ArXiv e-prints, 2012.
  • [36] G. Viennet. Inequalities for absolutely regular sequences: Application to density estimation. Probab. Theory Related Fields 107 (1997) 467–492.