Use the runs test to see if the sample series are distributed randomly.

Suppose you interviewed 30 people and want to see if the answers are random. (Answers that are not in random order may indicate a gradual bias in the phrasing of the questions or that the subjects are not being selected at random.)

The runs test is a nonparametric test because you make no assumptions about the distribution of the population being sampled.

The results of the runs test are valid if at least 10 observations are greater than the comparison criterion and at least 10 observations are less than or equal to the comparison criterion.

You want to compare the effects of three treatments using 35 subjects. You perform a randomness test to determine if the assignment of subjects to treatments was random. (To have valid results when applying a statistical test to compare the effects, the treatments must be randomly assigned to subjects.)

Data: Treatment.MTW (available in the Sample Data folder).