Revision history for PDL-Stats

0.3.0	12/03/2009
        GLM::anova_rptd! works for purely within designs
        GLM::anova bug fix--not messing up caller @ if passed @ ref in IV instead of pdl
	change prereq from PDL to PDL::Lite
	Stats.pm quick-start guide update
	t/stats_glm logistic test modified to use more proper numbers in x

0.2.8	11/04/2009
	fixed minor bug (mismatched nseed and self cluster pdls. oops) in switching to use cluster assignment as stop criteria. btw initial R2 now based on self instead of seeds.

0.2.7	11/04/2009
	GLM::plot_scree allows directly passing NCOMP. plots cutoff line for screes if requested
	kmeans stops when cluster assignment no longer changes instead of using R2CRT.
	if not specified, Kmeans::pca_cluster studies scree plot to determine NCOMP
	Basic::get_data STDERR message fix: data pdl o x v, NOT v x o

0.2.6	10/12/2009
        anova bad value se bug fix
	centroid bad value empty cluser bug fix
	basic, glm, distr rid of nan and use bad value instead

0.2.5   09/17/2009
        PREREQ_FATAL => 1 in Makefile.PL
        added pca_cluster in Kmeans.
        anova one-way bug fix.
        plot_means one-way bug fix.

0.2.4   09/02/2009
        added corr_table in PDL::Stats::Basic.
        PDL::Stats::GLM::pca uses corr_table.
        kmeans no longer quits with empty cluster. kmeans 3d threading bug fix.
        use short-circuit ISBAD(a) || ISBAD(b) instead of ISGOOD(a) && ISGOOD(b) in bad value processing.

0.2.3   08/24/2009
        ***anova interface change*** IV names now passed in \%opt instead of @ ref. I appologize for the interface change, but it matches the new method plot_means, and makes it easier to pass IVs themselves. I promise that such interface changes will be extremely rare, if at all.
        anova now supports bad value in DV.
        Added plot_means in PDL::Stats::GLM. anova plots highest order interaction by default.
        Added plot_scree in PDL::Stats::GLM. pca does scree plot by default.
        Added pca_sorti in PDL::Stats::GLM.
        plot_distr in PDL::Stats::Distr accepts WIN option
        Set prerequisite PDL => 2.4.4, which has bad value support by default

0.2.2   07/23/2009
        stats_distr.t moved form t/ to Distr/t, ie skip the test if no GSL.
        more kmeans fix (4d, seeding). added seeded kmeans tests.

0.2.1   07/22/2009
        kmeans 4d fix. added pms dir for cpan pod parse

0.2.0   07/21/2009
        kmeans threads! ss/ms bug fix. more tests on kmeans added.

0.1.3   07/17/2009
        quick-start for non-PDL people in Stats.pm pod
        ols_t uses double internally. GLM pod update
        r2_change threads (switched to use ols_t instead of ols)
        get_data which_id moved to Basic from GLM
        tests grouped in PDL-Stats/t except for PDL::GSL::CDF

0.1.2   07/15/2009
        Cleaned up some versioning stuff in Stats.pm
        GLM::get_data TYPE default to double. MISSN handling update; set to BAD by default.
        Case insensitive option passing.

0.1.1   07/14/2009
        Fixed PDL::Stats::GLM::ols b_p value for b_t < 0.

0.1.0   07/13/2009
        First version, released on an unsuspecting world.