Non-parametric Tests for Three or More Samples (Friedman and Kruskal-Wallis)

Abstract

The first part of this title contained all statistical tests relevant to starting clinical investigations, and included tests for continuous and binary data, power, sample size, multiple testing, variability, confounding, interaction, and reliability. The current part 2 of this title reviews methods for handling missing data, manipulated data, multiple confounders, predictions beyond observation, uncertainty of diagnostic tests, and the problems of outliers. Also robust tests, non-linear modeling , goodness of fit testing, Bhatacharya models, item response modeling, superiority testing, variability testing, binary partitioning for CART (classification and regression tree) methods, meta-analysis, and simple tests for incident analysis and unexpected observations at the workplace and reviewed. Each test method is reported together with (1) a data example from practice, (2) all steps to be taken using a scientific pocket calculator, and (3) the main results and their interpretation. Although several of the described methods can also be carried out with the help of statistical software, the latter procedure will be considerably slower.