Non-Parametric Analysis

Parametric calculations are necessary for analyzing data that is not continuous and/or non-normally distributed. Because these assumptions cannot be made, these tests are less powerful than parametric analysis. Non-parametric tests allow examination of data independent of the population distribution.

Rather than direct comparisons of the values in your fields, many non-parametric analyses rely on their relative rank. A wide variety of non-parametric tests are available and depend on the analysis you seek. There are single field (one sample), group comparisons, and paired fields analysis.

The first three options are for one sample test:

• Chi-Square — One sample Chi-Square: evaluates expected value for each unique value in a field.
• Sign Test — One sample sign test compares against the median, mean or user defined value.
• K-S Fit — Kolmogorov-Smirnov Goodness of Fit Test: compares against Uniform, Normal and Poisson distributions.

The following options compare groups of records:

• 2 Sample — Two sample tests: Wald-Wolfowitz Runs Test, Mann-Whitney U Test, and Kolmogorov-Smirnov.
• N Sample — Kruskal-Wallis One Way Analysis of Variance

The following options compare fields:

• Paired Fields — Field comparisons: paired sign test, Wilcoxon Signed Rank, Spearman's Correlation Coefficient.
• N Fields — Friedman's Two Way ANOVA of Ranks: Compare multiple fields with Friedman's two way ANOVA.