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



J.Q. Smith and Antonio Santos

Rolls Royce r

ime series

Sample 1

Sample 2

Sample 3

Mean

-0.0315

0.0126

0.0104

Variance

4.43597

3.7653

5.1998

Descriptive

Skewness

-0.9808

0.0239

-0.6057

Statistics

Kurtosis

11.2178

4.8352

14.8458

Min

-17.3518

-8.0899

-25.3876

Max

11.2178

8.8179

14.4814

Observations

1000

1000

1818

Stochastic

β

1.1405

1.1076

1.1535

Volatility

φ

0.9448

0.8358

0.9060

Estimates

ση

0.2078

0.3188

0.2784

Sample

First order

1

18

26

Impoverishment

Second order

0

0

0

Table 1: First section: Descriptive Statistics associated with each sub-sample of the series.
Second section: Estimates of the parameters for each sub-sample, which were obtained
as the mean of the posterior distribution defined through MCMC estimation techniques.
Third section: Sample impoverishment figures, which were calculated as the number of
times the range of the filter distribution was less than 0.2.

mean of the posterior distribution, which was obtained using Markov chain Monte
Carlo techniques. By considering three sub-samples we were able to analyse the
particle filter performance for different sets of parameters. A by-product of these
estimation procedures is the smoothing distribution of the states. Despite being
associated with different information sets, it is nonetheless convenient to compare
the smoothed distributions with the filter distributions obtained using algorithms
based on first and second order Taylor approximations to the log-likelihood. We ran

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

15




More intriguing information

1. Developing vocational practice in the jewelry sector through the incubation of a new ‘project-object’
2. The name is absent
3. Modelling the Effects of Public Support to Small Firms in the UK - Paradise Gained?
4. Impacts of Tourism and Fiscal Expenditure on Remote Islands in Japan: A Panel Data Analysis
5. Personal Income Tax Elasticity in Turkey: 1975-2005
6. The name is absent
7. The name is absent
8. The name is absent
9. TINKERING WITH VALUATION ESTIMATES: IS THERE A FUTURE FOR WILLINGNESS TO ACCEPT MEASURES?
10. Une Classe de Concepts