@Fuh: indeed it should and when I step carefully through the function, sometimes the result comes out correct and sometimes it doesn't, somewhat dependent on the numbers in alpha. To fix the problem go to lines 45 and 46 of circ_median (ver 2011f). You see two inequalities, dd>=0 and dd<0. The two inequalities should be identical for consistency and the correct result. Edit line 46 to read:
m2 = sum(dd<=0,1);
Now the function seems to behave as expected.
Um, what happened to my comment? What I wrote was that an typo error appears to have been introduced in circ_kuipertest.m in the advance to version 2011f. In 2010e, line 48 of the file reads:
[phis2 cdf2 phiplot2 cdfplot2] = circ_samplecdf(alpha2, res);
and in version 2011f, that line reads:
[~, cdf2 phiplot2 cdfplot2] = circ_samplecdf(alpha2, res);
and matlab complains of incorrect statement or expression.
Great submission. It would be nice to have cdf and inversion cdf for the vmpdf functions. Here's what I wrote for my needs
function p = circ_vmcdf(alpha, thetahat, kappa)
%integrates the pdf from an angle of -pi to an angle alpha
F = @(x)circ_vmpdf(x, thetahat, kappa);
p = quad(F,-pi(),alpha);
function theta = circ_vminv(p, thetahat, kappa)
%computes the inverse of the abovecirc_vmcdf.
fun =@(alpha)(circ_vmcdf(alpha, thetahat, kappa)-p);
theta = fzero(fun,[-pi pi]);