Should Normal Distribution be Normal? The Student's T Alternative

In the paper we try to answer, whether the Gaussian distribution - called widely the 'normal' distribution - is really basic, natural and normal. In particular, we investigate how the above statement conforms with the distribution of real data, namely daily returns of some stock indexes. It was the authors former experience that, when looking at the distributions of real data, it was very difficult to find there a 'normal', i.e. Gaussian distribution. The data, by their nature, are heterogeneous. If so, then the data should be modelled taking into account their possible heterogeneity. This can be done using mixture models - with mixtures composed from finite or infinite number of components. Students' T (univariate or multivariate) is one prominent example of distributions which may be obtained as a mixture of infinitesimal number of Gaussian distributions. The considerations are illustrated by an example of application to financial time series, namely daily returns of the indexes WIG20 and S&P500. We show, why the normality (i.e. 'Gaussianity') should be rejected and why the 't' distribution is plausible.

[1]  Paper ShawLee Copula Methods vs Canonical Multivariate Distributions: , 2007 .

[2]  K. Lange,et al.  Normal/Independent Distributions and Their Applications in Robust Regression , 1993 .

[3]  William T. Shaw,et al.  Sampling Student’s T distribution – use of the inverse cumulative distribution function , 2006 .

[4]  Geoffrey J. McLachlan,et al.  Finite Mixture Models , 2019, Annual Review of Statistics and Its Application.

[5]  Michel Verleysen,et al.  Flexible and Robust Bayesian Classification by Finite Mixture Models , 2004, ESANN.

[6]  W. Shaw Canonical Multivariate Distributions: the multivariate Student T distribution with general degrees of freedom , 2007 .

[7]  Anthony Brabazon,et al.  Biologically inspired algorithms for financial modelling , 2006, Natural computing series.

[8]  G. Booth,et al.  Statistical models of German stock returns , 1989 .

[9]  Radford M. Neal Pattern Recognition and Machine Learning , 2007, Technometrics.

[10]  Robert H. Kushler,et al.  Exploratory Data Analysis With MATLAB® , 2006, Technometrics.

[11]  Angel R. Martinez,et al.  : Exploratory data analysis with MATLAB ® , 2007 .

[12]  Christopher M. Bishop,et al.  Robust Bayesian Mixture Modelling , 2005, ESANN.

[13]  Jack C. Lee,et al.  Bayesian analysis of mixture modelling using the multivariate t distribution , 2004, Stat. Comput..

[14]  Nasser M. Nasrabadi,et al.  Pattern Recognition and Machine Learning , 2006, Technometrics.

[15]  P. Deb Finite Mixture Models , 2008 .

[16]  Michel Verleysen,et al.  Robust Bayesian clustering , 2007, Neural Networks.

[17]  Jeremy MG Taylor,et al.  Robust Statistical Modeling Using the t Distribution , 1989 .

[18]  V. Plerou,et al.  Scaling of the distribution of fluctuations of financial market indices. , 1999, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics.