On large deviations in testing Ornstein–Uhlenbeck-type models
Summary (1 min read)
1 Introduction
- Asymptotic properties of likelihood ratios play an important role in statistical testing problems.
- Likelihood ratio, Hellinger integral, Neyman-Pearson test, Bayes test, minimax test, large deviation theorems, Girsanov formula for diffusion-type processes, Ornstein-Uhlenbeck-type process, stochastic delay differential equation. in [4] were obtained by using the large deviation techniques for sequences of random variables, also known as Key words and phrases.
- The results are applied to the investigation of the rates of decrease for error probabilities of the tests mentioned above.
- Then the initial continuous-time assertions can be obtained as a limiting case of the corresponding results from the discrete-time case by applying the invariance principle.
2 Large deviation theorems and their applications
- In this section the authors cite some results from Lin’kov [21] - [23] and use his notation.
- In the rest of the section the authors refer some results about the asymptotic behavior of the error probabilities for Neyman-Pearson, Bayes, and minimax tests.
- The following assertion describes the rate of decrease for the error probabilities of the first and second kind αt and β(αt), respectively, for the test δt(αt) under the regularity condition (2.3).
- Under some other conditions the last equality in (2.19) was proved by Vajda [28].
3 Ornstein-Uhlenbeck models
- In this section the authors consider a model where the observation process X = (Xt)t≥0 satisfies the following stochastic differential equation: dXt = −θXt dt+.
- For this, the authors will investigate the asymptotic behavior of the Hellinger integral (3.5) under t→∞ .
- The question if the derived rate bounds are optimal as well as the second order expansions for log βt(αt) remain as open problems here.
4 Ornstein-Uhlenbeck-type models with delay
- Then, by means of the Girsanov-type formula (5.1) in [15], the authors get that under the hypothesis H0 the log-likelihood ratio process (2.1) admits the representation: Using the arguments in [21; Theorems 3.1.4, 3.2.2] they now describe the asymptotic behavior of the error probabilities for Neyman-Pearson tests.
- In the rest of the section the authors give some examples of models of the type (4.1) in which condition (4.7) holds.
- Financial support from the German Research Foundation and the Foundation of Berlin Parliament is gratefully acknowledged.
Did you find this useful? Give us your feedback
Citations
13 citations
12 citations
11 citations
Cites methods from "On large deviations in testing Orns..."
...To accomplish this, we will use appropriately Feynman–Kac formula (for a similar approach, see also Gapeev and Küchler [8])....
[...]
...[8] Pavel V. Gapeev and Uwe Küchler....
[...]
...To accomplish this, we will use appropriately Feynman–Kac formula (for a similar approach, see also Gapeev and Küchler [8])....
[...]
11 citations
7 citations
References
6,160 citations
Additional excerpts
...[29] or [5])....
[...]
5,987 citations
"On large deviations in testing Orns..." refers background in this paper
...for all t ≥ 0 (see e.g. Jacod and Shiryaev 1987; Chapt....
[...]
3,760 citations
"On large deviations in testing Orns..." refers background or methods in this paper
...Birg e [8] applied the results of [ 9 ] to the investigation of the rate of decrease for error probabilities of Neyman-Pearson tests....
[...]
...Cherno [ 9 ] proved large deviation theorems for sums of i.i.d....
[...]
...This assertion was proved by Cherno [ 9 ] for the case of i.i.d....
[...]
2,481 citations
"On large deviations in testing Orns..." refers methods in this paper
...H0 and H1, by means of the Girsanov formula for diffusion-type processes (see e.g. Liptser and Shiryaev 1977; Chapt....
[...]
...By applying Ito’s formula (see e.g. Liptser and Shiryaev 1977; Chapt....
[...]
2,113 citations
"On large deviations in testing Orns..." refers background in this paper
...From the formula (1.9.3) in Borodin and Salminen (1996) ; Chapt....
[...]
...(cf. the formula (1.9.3) in Borodin and Salminen 1996; Chapt....
[...]
...(cf. the formula (1.9.7) in Borodin and Salminen 1996; Chapt....
[...]
Related Papers (5)
Frequently Asked Questions (8)
Q2. What is the significance of the results of Lin’kov?
Lin’kov [22] proved large deviation theorems for extended random variables and applied them to the investigation of general statistical experiments.
Q3. What is the problem of testing hypotheses for affine delay differential equations?
Gushchin and Küchler [15] derived conditions under which a model with an affine stochastic delay differential equation satisfies the local asymptotic normality property and where the maximum likelihood and Bayesian estimators of a parameter are asymptotically normal and efficient.
Q4. What is the function lim t 1t logt = 0?
Then for the function ψt , t ≥ 0, given by:ψt = E0[ 12 ∫ t 0 Y 2s ds ] (4.5)we have lim t→∞ ψ−1t logαt = 0 implies lim sup t→∞ ψ−1t log β(αt) ≤ −1, (4.6)and if the condition
Q5. What is the simplest formula for the equation (3.1)?
by substituting the expression (3.7) into (2.3), taking ψt = θ1t and letting t go to ∞ , the authors get:κ(ε) = − √ ε(1− √ ε)2 and κ′(ε) = − 1 4 √ ε + 1 2 (3.8)for ε ∈ (ε−, ε+) = (0,∞), so that κ′(ε−+) = −∞ , κ′(1) = 1/4 and κ′(ε+−) = 1/2. It is easily seen that the function I(γ) from (2.5) takes the expression:I(γ) = sup ε>0 (εγ − κ(ε)) = 1 8(1− 2γ)(3.9)and the values in (2.7) - (2.8) can be calculated as γ0 = κ′(ε−+) = −∞ and γ1 = κ′(1) = 1/4 with I(γ0) = 0 and I(γ1) = 1/4.Because in this case the authors have γ0 < 0 < γ1 , from Propositions 2.1 - 2.4 and the formulas (3.8) - (3.9) the authors get that the following assertion holds.
Q6. How can the authors obtain the initial continuous-time assertions?
Then the initial continuous-time assertions can be obtained as a limiting case of the corresponding results from the discrete-time case by applying the invariance principle.
Q7. What are the properties of the delay test?
Asymptotic properties for tests of delay parameters in the cases of small noise and large sample size were recently studied by Kutoyants [19] - [20].
Q8. What is the simplest way to test the hypothesis?
Let αt , t ≥ 0, be the error probability of the first kind of the NeymanPearson test in the model (4.1) of testing hypothesis H0 : a(ds) ≡ a0(ds) against the alternative H1 : a(ds) ≡ a1(ds) with ai(ds) ∈