Rate of convergence for parametric estimation in a stochastic volatility model. (Q1766043)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Rate of convergence for parametric estimation in a stochastic volatility model.
scientific article

    Statements

    Rate of convergence for parametric estimation in a stochastic volatility model. (English)
    0 references
    0 references
    25 February 2005
    0 references
    The article investigates the rate of convergence of estimators \(\hat \theta \) of the finite-dimensional parameter \(\theta \) in the stochastic volatility equation \(v_t=v_0+\int _0^t\sigma (\theta , v_s)\,dW_s+\text{drift}\), observed via the diffusion process \(X_t=x_0+\int _0^tv_sdB_s+\text{drift}\) at times \(t=0,1/n,\dots,1\), where \((W_t)_{t\in [0,1]}\) and \((B_t)_{t\in [0,1]}\) are two independent Brownian motions. The final result of the article is a theorem stating that the rate of convergence \(\varphi _n =n^{-1/4}\) is a lower bound for \(\hat \theta \), in the sense \[ \liminf _{n\rightarrow \infty }\inf _{\hat \theta _n}\sup _{\text{drift}}\sup _\theta E^n_\theta \{\varphi ^{-2}_n| \hat \theta _n-\theta | ^2\}. \] Due to a recent result by \textit{A.\ Gloter} [C.\ R.\ Acad.\ Sci., Paris, Sér.\ I, Math.\ 330, 243-248 (2000; Zbl 0949.62073)], who exhibited a consistent \(n^{-1/4}\) estimator, \(\varphi _n =n^{-1/4}\) is actually an optimal rate of convergence. The theorem is first proved for the particular case of one-dimensional \(\theta \), zero drifts and \(\sigma (\theta , x)=\sqrt {\theta }\), to which the general case is then transformed. The proof follows a Bayesian approach of finding the least favourable prior, which is combined with the additional principle that bounds on priors are first established conditionally on elements of the Wiener space, and then transferred to the set of admissible values of \(\theta \) using the Wiener measure. It relies on numerous classical results of stochastic calculus, most heavily on the Burkholder-Davis-Gundy inequality, cf. \textit{D.L. Burkholder} and \textit{R.F. Gundy}, Acta Math. 124, 249--304 (1970; Zbl 0223.60021), as well as on a more recent result of \textit{B. Roynette} [Stochastics Stochastics Rep.\ 43, 221-260 (1993; Zbl 0808.60071)]. Although proving the above result is basically the only objective of the article, it is nevertheless composed of three differently oriented parts---an informal sketch of the main steps of the proof immediately after the formulation of the theorem, a mathematically rigorous proof of the theorem at the core of the article, and an appendix containing proofs of two technical lemmas. It is this structure that enabled the author to keep the technically complicated article transparent and, in spite of occasional missing or superfluous symbols in equations, well understandable.
    0 references
    stochastic volatility models
    0 references
    discrete sampling
    0 references
    high frequency data
    0 references
    nonparametric Bayesian estimation
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references