What Is the Difference Between Parametric and Non-Parametric Tests?
A parametric test is used on parametric data, while non-parametric data is examined with a non-parametric test. Parametric data is data that clusters around a particular point, with fewer outliers as the distance from that point increases. This data looks like a bell when graphed, with the particular point serving as its apex and the slopes moving downward in both directions to indicate reduced frequency. Non-parametric data does not focus on a particular point or appear this way when graphed.
Parametric data is considered to have a normal distribution because it tends to cluster at one specific point. This relationship is also described as a Gaussian distribution, after its discoverer, or simply as a bell curve. Since the shape is so predictable, it can be used to draw more reliable conclusions than comparatively random non-parametric data. Non-parametric data is less affected by extreme outliers and can be simpler to work with.
Parametric tests include the Pearson correlation test, independent-measures t-test, matched pair t-test and Anova tests. Non-parametric tests include the Spearman correlation test, Mann-Whitney test, Kruskal-Wallis test, Wilcoxon test and Friedman test. The number of data groups involved and the type of information desired dictates the best test to use, regardless of data type.
Tendencies of both data types can make it easier to tell which kind of a test is best suited to the situation. Parametric data tends to include ratios or intervals, while non-parametric data is either ordinal or nominal. The mean, or average, is the best measure of the midpoint of parametric data, while the median is more useful for non-parametric data. Non-parametric data can have any distribution or variance, while parametric data always has the homogeneous variance that produces a bell curve when graphed.