STAT_F_TEST.rar_F-test_The Test

function RT_SIG = STAT_F_TEST(PR_X,PR_Y,PR_X_MEAN,PR_Y_MEAN) %RT_SIG = STAT_F_TEST(PR_X,PR_Y,PR_X_MEAN,PR_Y_MEAN) % % This function performs a test for homoscedasticity (equal variance) % for two sets of sample data PR_X and PR_Y using the two sample % F-test. If the population means are known, it can be specified % in PR_X_MEAN and PR_Y_MEAN. If the population means are unknown, % leave these parameters unspecified or pass the empty matrix, []. % Do not simply pass the sample means. % % Parameters: % PR_X, PR_Y - a vector of samples from the two populations. % These vectors do not need to be the same length. % PR_X_MEAN, PR_Y_MEAN - the population means (scalar) from which the % samples were drawn, if known. if nargin < 3 PR_X_MEAN = []; end if nargin < 4 PR_Y_MEAN = []; end if isempty(PR_X_MEAN) LV_X_MEAN = mean(PR_X); if isreal(PR_X) LV_X_DOF = length(PR_X)-1; else LV_X_DOF = 2*(length(PR_X)-1); end else LV_X_MEAN = PR_X_MEAN; if isreal(PR_X) LV_X_DOF = length(PR_X); else LV_X_DOF = 2*length(PR_X); end end if isempty(PR_Y_MEAN) LV_Y_MEAN = mean(PR_Y); if isreal(PR_Y) LV_Y_DOF = length(PR_Y)-1; else LV_Y_DOF = 2*(length(PR_Y)-1); end else LV_Y_MEAN = PR_Y_MEAN; if isreal(PR_Y) LV_Y_DOF = length(PR_Y); else LV_Y_DOF = 2*length(PR_Y); end end % This ratio is F distributed if the samples are from a Gaussian populations with equal variances... LV_F_STAT = (LV_Y_DOF/LV_X_DOF)*(sum(abs(PR_X-LV_X_MEAN).^2)/sum(abs(PR_Y-LV_Y_MEAN).^2)); RT_SIG = fcdf(LV_F_STAT,LV_X_DOF,LV_Y_DOF); % The test can indicate that either the numerator or denominator of LV_F_STAT was too large to be expected. The significance % in a two-tailed test is twice the value of the least likely possibility (only one has likelihood less than 0.5). RT_SIG = 2*min(RT_SIG,1-RT_SIG);