# Ergodic and Prediction Problems

Chapter

## Summary

In this chapter we will be concerned with the asymptotic behavior of Markov processes with stationary transition probability function
or
hold for each set
or of
itself as

*P*(·,·). It is occasionally useful to look at the process only when it falls within a proper measurable subset*A*of the state space*Ω*. In the first section, conditions under which a well-defined “process on*A*” can be defined will be established. Sometimes it is useful to look at such derived processes when trying to establish the existence of*subinvariant*or*invariant*measures*μ*, that is,*σ*-finite measures for which$$ \int {\mu (dx)P(x,A) \leqslant \mu (A)} $$

$$ \int {\mu (dx)P(x,A) = \mu (A)} $$

*A*∈ A, respectively. The primary object of this chapter is the study of the asymptotic behavior of partial sums$$ \sum\limits_{j = 1}^n {{P_j}} ( \cdot , \cdot ) $$

(1)

$$ {P_n}( \cdot , \cdot ) $$

(2)

*n*→∞. Problems dealing with the partial sums (1) are related to ergodic theory while those concerned with the transition probability function (2) itself are more closely related to prediction. In discussions of ergodic and prediction problems, the existence of an invariant (or subinvariant) measure is usually required. Some conditions under which such a measure exists will be derived in section 3 under assumptions of a topological character. However, in discussions of ergodic and prediction problems the existence of such a measure will be taken for granted. In section 2 an ergodic theorem due to Chacon and Ornstein will be given with the classic Birkhoff ergodic theorem for stationary processes as an almost immediate corollary. Some aspects of a least squares version of the (nonlinear) prediction problem for a Markov process are sketched in section 4. It is quite clear that they are directly related to the asymptotic behavior of the transition function or equivalently to the asymptotic behavior of powers of the induced operator.## Keywords

Invariant Measure Ergodic Theorem Prediction Problem Compact Hausdorff Space Invariant Probability Measure
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

## Preview

Unable to display preview. Download preview PDF.

## Copyright information

© Springer-Verlag Berlin Heidelberg 1971