{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

rec08

# rec08 - MIT OpenCourseWare http/ocw.mit.edu 14.384 Time...

This preview shows pages 1–4. Sign up to view the full content.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: MIT OpenCourseWare http://ocw.mit.edu 14.384 Time Series Analysis Fall 2008 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms . More Empirical Process Theory 1 14.384 Time Series Analysis, Fall 2008 Recitation by Paul Schrimpf Supplementary to lectures given by Anna Mikusheva October 24, 2008 Recitation 8 More Empirical Process Theory This section of notes, especially the subsection on stochastic integrals are based on http://cscs.umich. edu/ ~ crshalizi/weblog/472.html . Convergence of Random Walks 1 [ τT ] We never did formally show the weak convergence of to a Brownian motion. Let’s fill that hole √ T t =1 now. Remember the functional central limit theorem. Theorem 1. Functional Central Limit Theorem : If 1. there exists a finite-dimensional distribution convergence of ξ T to ξ (as in ( ?? )) 2. ∀ ε,η > there exists a partition of Θ into finitely many sets, Θ 1 ,... Θ k such that lim sup P (max sup ξ T ( τ 1 ) − ξ T ( τ 2 ) > η ) < ε T →∞ i τ 1 ,τ 2 ∈ Θ i | | then ξ T ξ ⇒ Finite dimensional convergence follows from a standard central limit theorem. We just need to verify the second condition, stochastic equicontinuity. It is not very hard to do this directly. Θ = [0 , 1] is compact, so we can choose our partition to be a collection of intervals of length δ . This gives P (max sup ξ T ( τ 1 ) − ξ T ( τ 2 ) > η ) ≤ P ( sup ξ T ( τ 1 ) − ξ T ( τ 2 ) > η ) i τ 1 ,τ 2 ∈ Θ i | | | τ 1 − τ 2 | <δ | | 1 [( τ + δ ) T ] ≤ P ( sup t ) τ ∈ [0 , 1] |√ T | t =[ τT ] Chebyshev’s inequality says that if E [ x ] = µ and V ar ( x ) = σ 2 , then P ( x − µ > η ) ≤ σ η 2 2 . Here, [( τ + δ ) T ] [( τ + δ ) T ] | | 1 1 E [ √ T t =[ τT ] t ] = 0 and V ar ( √ T t =[ τT ] t ) = [ δT T ] σ 2 and these things do not depend on τ , so [( τ + δ ) T ] P ( sup 1 t > η ) ≤ [ δT ] σ 2 |√ T | Tη 2 τ ∈ [0 , 1] t =[ τT ] Hence, setting δ ≤ η 2 ε we have the desired result. σ 2 Stochastic Integrals 2 Stochastic Integrals Our interest in stochastic integrals comes from the fact that we would like to say something about the convergence of 1 1 y t t = ξ T ( t/T ) √ T ( ξ T ( t/T ) − ξ T (( t − 1) /T )) T T We know that ξ T ⇒ W , a Brownian motion. If W were differentiable, we’d also know that lim T →∞ √ T ( W ( τT/T ) − W (([ τT ] − 1) /T )) = W ( τ ). Finally, a sum like the above would usually converge to an integral T 1 1 f ( t/T ) g ( t/T ) f ( x ) dg ( x ) T → t =1 We would like to generalize the idea of an integral to apply to random functions such as ξ . There are a few problems. First, W is not differentiable. Nonetheless, we can still look at T 1 lim ξ ( t/T )( W ( t/T ) − W (( t − 1) /T )) T →∞ T t =1 We must verify that this sum converges to something. We will call its limit a stochastic integral, and write 1 it as ξ ( t ) dW ( t ). We then want to define stochastic integrals...
View Full Document

{[ snackBarMessage ]}

### Page1 / 6

rec08 - MIT OpenCourseWare http/ocw.mit.edu 14.384 Time...

This preview shows document pages 1 - 4. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online