| Abstract:
The opportunity for descriptive and experimental inquiry using qualitative evaluation data is often overlooked due to concerns about small samples, ordinal data, or violations of normality that common parametric statistical tests rely upon. Nonparametric methods can be more powerful than parametric methods if the assumptions behind the parametric model do not hold. We will explore a number of practical applications of common nonparametric analysis methods that are appropriate for counts, ordered-categorical, non-ordered categorical, small samples, data for one or several groups of subjects, and data collected at multiple time points. Common nonparametric tests will be discussed including Mann-Whitney, Wilcoxon, Kruskall-Wallis, and Freidman as well as common nonparametric correlation coefficients including Spearman R, Kendall Tau, and Gamma. We will review the rationale and assumptions underlying each method and discuss their weaknesses and strengths. Examples of applications in program evaluation will be used to illustrate each method.
|