Dependent variable Number Measurement level Nonmetric Binary Counts

Dependent variable Number Measurement level* Non-metric Binary Counts 1 Nominal Ordinal Metric Independent variable Number 1 1 1 Analysis method Measurement level Non-metric Nonparametric tests, χ², logistic analysis of variance, loglinear analysis Binary Mann-Whitney test, χ² Metric Logistic regression analysis Non-metric χ², Cramér’s coefficient C Metric χ², logistic regression analysis Non-Metric χ² Metric Regression Non-metric Cramér’s coefficient C Metric (Logistic) regression analysis, loglikelihood Binary Wilcoxon’s two sample test, χ², Mann-Whitney test, Kolmogorov-Smirnov test Counts Nonparametric correlation Nominal Kruskal-Wallis test, analysis of variance Ordinal Nonparametric correlation, Spearman’s rho, Kendall’s tau Metric Nonparametric correlation, isotonic regression analysis Binary Correlation, t-test Counts Regression analysis Nominal Analysis of variance Ordinal Nonparametric correlation Metric Regression analysis, correlation

Dependent variable Number Measurement level* Non-metric 1 Independent variable Number Measurement level Non-metric Logistic analysis of variance, loglinear analysis , canonical correlation analysis Metric Logistic regression analysis Metric Discriminant analysis Non-metric Analysis of varaince Metric Multiple regression analysis Non-metric Multivariate regression with dummy variables Non-metric Multivariate analysis of variance with dummy variables Metric Multivariate multiple regression with dummy variables Non-metric Multivariate analysis of variance Metric Multivariate regression analysis Non-metric Multivariate analysis of variance Metric Multivariate multiple regression analysis, Canonical correlation analysis, redundancy analysis 2 or more Nominal 2 or more Metric 2 or more 1 Non-metric 2 or more 1 Metric Analysis method 2 or more

Number of Measurement variables level* Analysis methods Description Confirmation Binary Frequencies, proportions, counts Nonparametric tests, binomial test Counts Frequencies, mean, median, mode Nonparametric tests, t-test Nominal Frequencies, proportions, mode Nonparametric tests, χ² Ordinal Frequencies, mean, median, mode Nonparametric tests, binomial test Metric Mean, median, mode, variance, kurtosis Confidence intervals, t-test Binary Tetrachoric correlation, crosstabulations Nonparametric tests Counts Crosstabulations Nonparametric tests, χ² Nominal Contingency tables, correspondence analysis Loglinear analysis, χ² Ordinal Kendall’s tau, Spearman’s rho Nonparametric tests Metric Correlation coefficient, scatter plot Correlation coefficient Binary Rasch models, Guttman scaling Loglinear analysis Counts Crosstabulations Factor analysis Ordinal Nonlinear principal component analysis, multidimensional scaling Factor analysis Metric Principal component analysis, one-dimensional scaling, cluster analysis Factor analysis, F-test Nominal Latent class analysis, contingency tables, correspondence analysis, homogeneity analysis Loglinear analysis 4 or 5 Nominal Latent class analysis, homogeneity analysis Loglinear analysis 6 or more Nominal Homogeneity analysis Loglinear analysis 1 2 More than 2 3

• • Analysis methods suitable for issues with both dependent en independent variables. Derived from Van den Berg (1992, Appendix) Analysis methods for issues without distinction between dependent en independent variables. Derived from Van den Berg (1992, Appendix) * In this diagram, binary variables, counts, nominal variables and ordinal variables have been distinguished in addition to the rough distinction between metric (or numerical) and non-metric (or categorical) variables. Ordinal variables are most often treated like metric variables, if their distributions are normal. However, there (often) are special analysis methods for cases in which ordinal variables are involved. Binary variables, nominal variables and counts are often considered as non-metric variables, so if there is no analysis method which is designed specifically for one of these measurement levels, please use the method suitable for a non-metric variable. (See also Van de Berg, 1992, pages 31 and 32). Gerda van den Berg (1992). Choosing an Analysis Method: An Empirical Study of Statisticians' Ideas in View of the Design of Computerized Support. Leiden: DSWO Press (ISBN 9090046348)