Abstract
This paper contributes to approximate analysis of dynamic generalized linear models (DGLM). The issue is approximation of predictive integrals for DGLM. Such integrals arise e.g. when computing the likelihood function of data conditional on a fixed hyperparameter. Recall that in dynamic linear modelling it is assumed that each observation y t of a (univariate) time series y 1,...,y N is a realization from a distribution of the exponential family with parameters changing over time ([14]). This change is explained through an unobservable state vector x t ∈ ℝr which follows a Gaussian process. Conditional on a known value of x t the distribution of y t |x t is assumed to belong to the exponential family. By predictive integral we now mean integrals of the type
with t > s and \(y^s = \{y_i \cdots, y_s\}\). p(y t|y s) and p(y t|x t) denote the densities of the distribution of y t|y s and y t |x t with respect to the Lebesgues (metric time series) or to the counting measure (discrete time series). Integrals of this type can be thought of as an infinite mixture of conditional (‘mixed’) densities with a mixing measure P s which has ‘mixing’ Lebesgue density \(p^s(\textbf{\textit{x}}_t)= p(\textbf{\textit{x}}_t|y^s)\):
The problem of approximation of predictive integrals is most often discussed for i.i.d. observations arising from models with static parameters x t = x (e.g. [8], [11], [6]). The methods suggested in these papers approximate the infinite mixture (2) by a finite number of mixed densities where the choice of the conditional parameter is based on the mixing density:
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Kitagawa, G.: Non-Gaussian State Space Modelling of Nonstationary Time Series. JASA, 82, 1987, 1032–1063.
Fahrmeir, L.: Posterior mode estimation by extended Kalman-filtering for multivariate dynamic generalized linear models. To appear in JASA.
Fahrmeir, L. and Kaufmann, H.: On Kaiman-Filtering, Posterior Mode Estimation and Fisher-Scoring in Dynamic Exponential Family Regression. Metrika, 1991, 38, 37–60.
Frühwirth-Schnatter, S.: Monitoring von ökologischen und biometrischen Prozessen mit statistischen Filtern. In Seeber, G.U.H. and Minder, Ch. (Eds.): Multivariate Modelle, 89–122. Springer, 1991.
Harrison, P.J. and Stevens, C.F.: Bayesian Forecasting. J. R. Stat.Soc., B, 38, 1976, 205–247.
Kass, R.E., Tierney, L. and Kadane, J.B.: Fully Exponential Laplace Approximations to Expectations and Variances of non-positive Functions. JASA, 84, 407, 1989, 710–716.
Magill, D.T.: Optimal Adaptive Estimation for Sampled Stochastic Processes. IEEE-TAC, 10, 1965, 434–439.
Naylor, J.C. and Smith, A.F.M.: Application of a Method for the Efficient Computation of Posterior Distributions. Applied Statistics, 31, 1982, 214–225.
Schnatter, S.: Bayesian Forecasting of Time Series Using Gaussian Sum Approximation. In: J.M. Bernardo, M.H. DeGroot, D.V. Lindley and Smith A.F.M. (Eds.), Bayesian Statistics 3, 757–764. University Press, Oxford, 1988.
Schnatter, S.: Integration-based Kalman-filtering for a dynamic generalized linear trend model. To appear in Computational Statistics and Data Analysis, 13, 1992.
Tanner, M: Tools for Statistical Inference. Observed Data and Data Augmentation Methods. Lecture Notes in Statistics, 67. Springer, New York, 1991.
Thompson, P.A. and Miller, R.B.: Sampling the Future — a Bayesian Approach to Forecasting. Journal of Business and Economic Statistics, 4, 1986, 427–436.
West, M.: Bayesian computations — sequential analysis and dynamic models. ISDS Discussion Paper #90-A12, Duke University, 1990.
West, M. and Harrison, P.J.: Bayesian Forecasting and Dynamic Models. Springer, New York/Heidelberg/Berlin, 1989.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1992 Springer-Verlag New York, Inc.
About this paper
Cite this paper
Frühwirth-Schnatter, S. (1992). Approximate Predictive Integrals for Dynamic Generalized Linear Models. In: Fahrmeir, L., Francis, B., Gilchrist, R., Tutz, G. (eds) Advances in GLIM and Statistical Modelling. Lecture Notes in Statistics, vol 78. Springer, New York, NY. https://doi.org/10.1007/978-1-4612-2952-0_16
Download citation
DOI: https://doi.org/10.1007/978-1-4612-2952-0_16
Publisher Name: Springer, New York, NY
Print ISBN: 978-0-387-97873-4
Online ISBN: 978-1-4612-2952-0
eBook Packages: Springer Book Archive