|
|
Readme file
SERIES
B
 Statistical
Methodology
J. Fortiana and A. Grané: Goodness-of-it tests
based on maximum correlations and their orthogonal decompositions
Journal of the Royal Statistical Society, Series B, volume 65 (2003),
part 1, pages 115 - 126
gof.m
is an Octave/Matlab program which implements the proposed test for
a given sample, given as a vector x.
[qn,answer,lower,upper]=gof(x)
implements a two-tail 5% significance test of uniformity based on the
statistic Qn. H0 is the null hypothesis that the sample is
uniformly distributed in [0,1]. For a sample size of 2<=n<=50
the exact test is performed, while for n>50 the asymptotic test is
performed.
The output is as follows:
qn is the computed value of the test satistic, answer is a character
string: either "accept H0" or "reject H0", lower
is the lower-tail critical value, upper is the upper-tail critical value.
gof.nb
is a Mathematica-Compatible Notebook, which can be used on any computer
system with Mathematica 3.0 (or higher). It contains functions to compute
the exact probability density function of the L-statistic Q_n, following
the algorithm described in the paper and references, and critical values
for the test. These computations involve both symbolic and numerical
computations. The adjustable parameter Ndec gives the number of digits
to use in the latter. It has to be rather large in order to allow for
the loss of precision associated with solving polynomial equations of
a possibly large degree.
J. Fortiana
Departament d'Estadistica
Facultat de Matematiques
Universitat de Barcelona
Gran Via de les Corts Catalanes 585
08007-Barcelona
Spain
E-mail: fortiana@mat.ub.es
|
Journals
SERIES A
Statistics in Society
SERIES B
Statistical
Methodology
SERIES C
Applied Statistics
SERIES D
The
Statistician
|
