Творчі рішення як похідні від вибіркового множинного тестування

Abstract

Algorithm of creative solutions as derivatives of selective multiple testing: A. Initial selection of multiple testing methods A1. Selection of independent and dependent variability; Calculating the of mean, standard error of mean, standard deviation, 95% confidence interval for mean, median, minimum, maximum, range, quartiles; Determination of the variabilities distribution - parametric or nonparametric by single-factor the Kolmogorov-Smirnov test; Shapiro-Wilk W test and graphical methods: frequency distribution histograms stem & leaf plots; scatter plots; box & whisker plots; normal probability plots: PP and QQ plots; graphs with error bars (Graphs: Error Bar). A2. ANOVA (Analysis of Variance) test is used for parametric variabilities distribution. If deviations are homogeneous by Levene test would used the method of multiple comparison groups by Tukey HSD, Scheffe, Bonferroni, and in the cases without homogeneity we must use the criteria Tamhane's T2, Games-Howell; Kruskal-Wallis test, nonparametric equivalent of the ANOVA, is used for nonparametric variabilities distribution; A3. The selection of variabilities, as criteria for making decisions, with P = .05 or less, and / or minimal false discovery rate, q-value (Gyorffy B, Gyorffy A, Tulassay Z:. The problem of multiple testing and its solutions for genom-wide studies. Orv Hetil, 2005;146(12):559-563) Determination of the sensitivity and specificity of these variabilities. B. Secondary screening the variabilities for multiple test methods. B1. These numerical dependent variabilities with P = .05 or less, and / or minimal false discovery rate, with high sensitivity and specificity by diagnostic capabilities must use for formation of new variabilities as descendants of 2, 3, 4 .. n numerical dependent variabilities as the derivatives of various mathematical transformations as Cantor, Sierpinski, von Koch sets, etc., anti-fractal sets; Moebius strip like aggregates, oxymoron combinations (Kulishov S.K., Iakovenko O.M.: Fractal and antifractal oxymorons, Moebius strip like transformations of biomedical data as basis for exploratory subgroup analysis. Book of abstract of International Conference on Trends and Perspective in Linear Statistical Inference; LinStat, 2014, Linkoping, Sweden, August 24-28, 2014; 2014, 58); and others mathematical transformations derivatives. C. Check the newly formed variabilities similar to step A to estimate the effectiveness of such changes. D. Comparison of multiple testing of more informative primary and secondary variabilities by accuracy, sensitivity and specificity of diagnostic possibilities. E. If it's necessary, the search of new selection principles of variabilities for multiple testing must be continued.