Robust estimation of crosscovariance function in the presence of outliers for large sample series

Abstract

This paper considers the effect of outliers on cross covariance as model identification and specification tool. We established that outliers in series significantly affect the mean of cross-covariance function (CCF). For large samples, the asymptotic convergence of cross-covariance function expectations are infested if the original series is classified as 2-dimensional random fields in the presence of outliers. Robust estimates of cross-covariance function that accommodate outliers are proposed. Analytically, the proposed Jacknife (JK) estimate reduced the bias in the conventional method (CM) of estimation by 50%. When the upper bound of the outlier infested part of JK estimate is constrained to the number of outliers in the series, the new Jacknife estimate (NJK) completely removes the outlier infested bias in CM. The empirical illustration with real-life data showed that NJK estimates have lowest standard errors compared with CM and JK and decay with increase in lag.

Keywords

Outlying observations transfer function crosscovariance modified jackknife estimate

References

Box,

G E P

and Jenkins,

G M

(1976): Time series analysis. Forecasting and Control. New Jersey: Prentice Hall.

Chernick,

M R

, Downing,

D J

, and Pike,

D H

(1982): Detecting outliers in time series data. Journal of American Statistical Association, 77, 380.

Choi,

(2007): The asymptotic joint distribution of the Yule-Walker estimators of a causal multidimensional AR process. Communication in Statistics Theory Method, 30(3) pp 609-614.

Gomez,

(2009): Transfer function model Identification. Boletin de Estadistica e Investigacion Operativa, 25, 2, 109-115.

Kudelas,

(1994): The linear model variance-covariance components and jackknife estimation. Application of mathematics, 39, 2, 111-125.

Liu,

L M

and Hanssens,

D M

(1982): Identification of multiple-input transfer function models. Communication in Statistics. Theory and Methods, 11, 297-314.

Nogales,

F J

and Conejo,

A J

(2006): Electricity price forecasting through transfer function models. Journal of the operational research society, 57, 350-356.

Rollins

D K

, Bhandari

, Chin

, Junge

T M

, and Roosa

K M

(2006): Optimal deterministic transfer function modeling in the presence of serially correlated noise. Chemical engineering research and design, 84(A1), 9-21.

Shangodoyin,

D K

(1994): On the specification of time series models in the presence of aberrant observations. Unpublished PhD Thesis. University of Ibadan, Ibadan, Nigeria.

10.

Shangodoyin,

D K

and Arnab,

(2005): The Computational efficiency of Outlier estimates for different generating process. Model Assisted Statistics and Applications, 1, # 2, 69-75. IOS Press Amsterdam. Statistics.

11.

Tsay,

R S

(1985): Model identification in Dynamic Regression models. Journal of Business and Economic Statistics, 3, 228-237.

12.

Tsay,

R S

(1986): Time series model specification in the presence of outliers. Journal of American Statistical Association, 81, 132-141.

13.

Wei,

W S

(2006): Time series analysis (univariate and Multivariate methods). Pearson-Addison Wesley.