TITLE:
A Look at Sequential Normal Scores and How They Apply to Financial Data Analysis
AUTHORS:
W. J. Conover, Víctor G. Tercero-Gόmez, Alvaro E. Cordero-Franco
KEYWORDS:
Sequential Analysis, Sequential Ranks, Transformation, Extreme Values, Outliers
JOURNAL NAME:
Journal of Applied Mathematics and Physics,
Vol.6 No.4,
April
24,
2018
ABSTRACT: Statistical methods for analyzing economic data need
to be timely, accurate and easy to compute. To accomplish this, parametric models are often assumed, but they are at best approximate, and often lack a good
fit in the tails of the distribution where much of the interesting data are concentrated. Therefore, nonparametric methods
have been extensively examined as alternatives to the constrictive assumptions
of parametric models. This paper examines the use of Sequential Normal Scores
(SNS) for transforming time series data with unknown distributions into time
series data that are approximately standard-normally distributed. Particular
attention is directed toward detecting outliers (out-of-control values), and
applying subsequent analytic methods such as CUSUMs and Exponentially Weighted
Moving Average (EWMA) schemes. Two examples of stock market data are presented
for illustration.