function screeplot(vec,attrib,lev)
%SCREEPLOT draws the eigenvalues of the covariance matrix of the data in decreasing order.
% It can be used to decide how many latent variables to retain in a PCA analysis
% by looking at the kink in the curve.
%
% Reference: J.A. Jackson (1991), "A User's Guide to Principal Components",
% Wiley Series in Probability and Mathematical Statistics, New-York.
%
% I/O: screeplot(vec,attrib,lev)
%
% Required input arguments:
% vec : eigenvalues
% attrib : class identifier
%
% Optional input argument:
% lev : 0 (default) or 1 to include a plot of the logarithm of the eigenvalues
%
% Created 28 July 2000 by Sabine Verboven
% Last Update: 20/06/2003
if nargin < 3
lev=0;
end
if nargin < 2
error('no class identifier is given.')
end
%%%%%%%MAIN%%%%%%%
set(gcf,'Name', 'Scree plot','NumberTitle', 'off');
cla
hold on
ymin=max([0,min(vec)]);
ymax=max(vec);
ymarg=0.06*(ymax-ymin);
ymin=ymin-ymarg;
ymax=ymax+ymarg;
plot(vec,'o-');
xlabel('Index')
ylabel('Eigenvalue')
if length(vec)==1
xlim([0,2]);
ymarg=0.5;
ymin=ymin-ymarg;
ymax=ymin+(2*ymarg);
else
xlim([1,length(vec)]);
end
ylim([ymin,ymax]);
if lev==1
set(gcf,'Name', 'LEV plot','NumberTitle', 'off');
hold on
plot(log(vec),'b.-');
xlabel('Number of LV')
ylabel('Log(eigenvalue)')
end
box on
title(attrib)
hold off