Second Order Filter Distribution Approximations for Financial Time Series with Extreme Outlier



J.Q. Smith and Antonio Santos

the following equations,


f (αt,klDt)   α  f (ytat) f (αtαt-1,k)                               (13)

≤  g (yt↑αt,μt,k)f (atat-1,k)                         (14)

=  g yytμ^k,k) g a*tat-1,k, yt, μt,k)                  (15)


a g (at,kDt)                                   (16)


To implement rejection sampling, k is first sampled from a density proportional
to
g (ytμt,k}, then at is drawn from g (atat-1 ,k,ytt,k), which is equivalent of


sampling from g (αt, k|Dt).


As this is an approximating density, the pair (k, αt) is


accepted with probability


f (at,kDt)
g (at, kDt)


(17)


which can be rewritten as


f (ytat)
g yytat,μkk)


(18)


Pitt and Shephard (1999, 2001) developed these results and applied them to the

SV model in (1)-(2). They used μt,k = φαt-1 ,k and the first order approximation


log g (ytat,μt,k) = const


at

2


yt2


2β2 exp μkttk)


μt,k         (19)


which, combined with the Gaussian transition density, gives rise to an approximating
density that can be factorized into two densities

g (αtαL1,ytl,k) = N μ'k-σ.2)

(20)


and


g yt μtkk a exp


μμt,k μt,k f yt (1 + μt,k)
tʒfe exp   2β2 exp (μtkk)


where


„.._,,.+ ⅛(____y2__1

μtk   μtk +2 ' 2 exp (μt,k)


(21)

(22)


G.E.M.F - F.E.U.C.



More intriguing information

1. The name is absent
2. The name is absent
3. Moffett and rhetoric
4. Psychological Aspects of Market Crashes
5. Correlates of Alcoholic Blackout Experience
6. Reconsidering the value of pupil attitudes to studying post-16: a caution for Paul Croll
7. The name is absent
8. Economies of Size for Conventional Tillage and No-till Wheat Production
9. Non-causality in Bivariate Binary Panel Data
10. Crime as a Social Cost of Poverty and Inequality: A Review Focusing on Developing Countries