epsilon

Corrects for violations of sphericity in repeated-measures ANOVA designs

% [epsilonhat DFn DFd] = epsilon (y, subj, gp, alg, DFn, DFd)
%
% Repeated-measures ANOVA assumes that error is purely random.
% A random factor that causes a measurement in one subject
% to be a higher (or lower) should have no affect on the next
% measurement in the same subject. This assumption is called
% circularity or sphericity. Sphericity, is therefore a measure
% of equality of variances of the differences between treatment
% levels.
%
% The adjusted degrees of freedom is returned to correct the p-
% value obtained from the treatment F statistic:
%
% padj = 1-fcdf(F,DFn,DFd)
%
% y is a vector of the data
% subj is a vector of subject identifiers
% gp is the number of treatment groups
% alg is the algorithm used (see below)
% DFn is the degrees of freedom in the numerator
% DFd is the degrees of freedom in the numerator
%
% y values must be ordered appropriately for each subject
%
% The algorithms include:
% 1) Greenhouse-Geisser (GG)
% - for 1-way repeated-measures ANOVA (conservative)
% 2) Huynh-Feldt-Lecoutre correction (HFL)
% - for 1-way repeated-measures ANOVA (less conservative)
% - for correcting violations of multisample sphericity in
% repeated-measures designs with >=2 independent groups
%
% The default is 'HFL', corresponding to the Lecoutre corrected form of
% the Huynh-Feldt adjustment [3]. The alternative is 'GG' for the
% Greenhouse-Geisser adjustment.
%
% Bibliography:
% [1] Maxwell and Delaney (2004) Designing Experiments and Analyzing
% Data: A Model Comparison. Psychology Press. Vol 1 p543
% [2] http://homepages.gold.ac.uk/aphome/spheric.html
% [3] Keselman, Algina and Kowalchuk (2001) The analysis of repeated
% measures designs: A review. British Journal of Mathematical
% and Statistical Psychology (2001), 54, 1-20