statistics.py 文件源码

def main(pkl_list, name_list, cut=sys.maxint):
    pickles = plot_util.load_pickles(name_list, pkl_list)
    best_dict, idx_dict, keys = plot_util.get_best_dict(name_list, pickles,
                                                       cut=cut)

    for k in keys:
        sys.stdout.write("%10s: %s experiment(s)\n" % (k, len(best_dict[k])))

    sys.stdout.write("Unpaired t-tests-----------------------------------------------------\n")
    # TODO: replace by itertools
    for idx, k in enumerate(keys):
        if len(keys) > 1:
            for j in keys[idx+1:]:
                t_true, p_true = stats.ttest_ind(best_dict[k], best_dict[j])
                rounded_t_true, rounded_p_true = stats.ttest_ind(numpy.round(best_dict[k], 3),
                                                                 numpy.round(best_dict[j], 3))

                sys.stdout.write("%10s vs %10s\n" % (k, j))
                sys.stdout.write("Standard independent 2 sample test, equal population variance\n")
                sys.stdout.write(" "*24 + "  T: %10.5e, p-value: %10.5e (%5.3f%%) \n" %
                                (t_true, p_true, p_true*100))
                sys.stdout.write("Rounded:                ")
                sys.stdout.write("  T: %10.5e, p-value: %10.5e (%5.3f%%)\n" %
                                (rounded_t_true, rounded_p_true, rounded_p_true*100))
                if tuple(map(int, (scipy.__version__.split(".")))) >= (0, 11, 0):
                    # print scipy.__version__ >= '0.11.0'
                    t_false, p_false = stats.ttest_ind(best_dict[k], best_dict[j], equal_var=False)
                    rounded_t_false, rounded_p_false = stats.ttest_ind(numpy.round(best_dict[k], 3),
                                                                       numpy.round(best_dict[j], 3),
                                                                       equal_var=False)
                    sys.stdout.write("Welch's t-test, no equal population variance\n")
                    sys.stdout.write(" "*24)
                    sys.stdout.write(": T: %10.5e, p-value: %10.5e (%5.3f%%)\n" %
                                    (t_false, p_false, p_false*100))
                    sys.stdout.write("Rounded:                ")
                    sys.stdout.write(": T: %10.5e, p-value: %10.5e (%5.3f%%)\n" %
                                    (rounded_t_false, rounded_p_false, rounded_p_false*100))
                sys.stdout.write("\n")

    sys.stdout.write("Best Value-----------------------------------------------------------\n")
    for k in keys:
        sys.stdout.write("%10s: %10.5f (min: %10.5f, max: %10.5f, std: %5.3f)\n" %
                        (k, float(numpy.mean(best_dict[k])), float(numpy.min(best_dict[k])),
                         numpy.max(best_dict[k]), float(numpy.std(best_dict[k]))))

    sys.stdout.write("Needed Trials--------------------------------------------------------\n")
    for k in keys:
        sys.stdout.write("%10s: %10.5f (min: %10.5f, max: %10.5f, std: %5.3f)\n" %
                        (k, float(numpy.mean(idx_dict[k])), float(numpy.min(idx_dict[k])),
                         numpy.max(idx_dict[k]), float(numpy.std(idx_dict[k]))))

    sys.stdout.write("------------------------------------------------------------------------\n")