What is the Mann Whitney Test?
The Mann Whitney test (also called Mann–Whitney U test or Wilcoxon rank-sum test) is a statistical hypothesis test to compare the medians of two populations that are not normally distributed. In a non-normal distribution, the median is the better representation of the center of the distribution.
- Null Hypothesis (H0): η1 = η2
- Alternative Hypothesis (Ha): η1 ≠ η2
Where:
- η1 is the median of one population
- η2 is the median of the other population
- The null hypothesis is that the medians are equal, and the alternative is that they are not equal
Mann Whitney Test Assumptions
- The sample data drawn from the populations of interest are unbiased and representative.
- The data of both populations are continuous or ordinal when the spacing between adjacent values is not constant. (Reminder: Ordinal data—A set of data is said to be ordinal if the values can be ranked or have a rating scale attached. You can count and order, but not measure, ordinal data.)
- The two populations are independent to each other.
- The Mann–Whitney test is robust for the non-normally distributed population.
- The Mann–Whitney test can be used when shapes of the two populations’ distributions are different.
How Mann Whitney Test Works
Step 1:
Group the two samples from two populations (sample 1 is from population 1 and sample 2 is from population 2) into a single data set and then sort the data in ascending order ranked from 1 to n, where n is the total number of observations.
Step 2:
Add up the ranks for all the observations from sample 1 and call it R1. Add up the ranks for all the observations from sample 2 and call it R2.
Step 3:
Calculate the test statistics
Where:
and where:
- η1 and η2 are the sample sizes
- R1 and R2 are the sum of ranks for observations from sample 1 and 2 respectively
Step 4:
Make a decision on whether to reject the null hypothesis.
- Null Hypothesis (H0): η1 = η2
- Alternative Hypothesis (Ha): η1 ≠ η2
If both of the sample sizes are smaller than 10, the distribution of U under the null hypothesis is tabulated.
- The test statistic is U and, by using the Mann–Whitney table, we would find the p-value.
- If the p-value is smaller than alpha level (0.05), we reject the null hypothesis.
- If the p-value is greater than alpha level (0.05), we fail to reject the null hypothesis.
- If both sample sizes are greater than 10, the distribution of U can be approximated by a normal distribution. In other words, (U-μ)/σ follows a standard normal distribution.
Z_calc=(U-μ)/σ
Where:
μ=(η1 η2)/2
σ=√(√(η1 n_2 (η1+η2+1))/12)
If the sample sizes are greater than 10, then the distribution of U can be approximated by a normal distribution. The U value is then plugged into the formula seen here to calculate a Z statistic.
When |Zcalc| is greater than Z value at α/2 level (e.g. when α = 5%, the z value we compare |Zcalc| to is 1.96), we reject the null hypothesis.
Use Minitab to Run a Mann–Whitney Test
Case study: We are interested in comparing customer satisfaction between two types of customers using a nonparametric (i.e. distribution-free) hypothesis test: Mann–Whitney test.
Data File: “Mann–Whitney” tab in “Sample Data.xlsx”
- Null Hypothesis (H0): η1 = η2
- Alternative Hypothesis (Ha): η1 ≠ η2
Steps to run a Mann–Whitney Test in Minitab:
- Un-stack the data into two separate columns:
- “Overall Satisfaction 1” for customer type = 1
- “Overall Satisfaction 2” for customer type = 2
- Click Stat → Nonparametrics → Mann–Whitney.
- A new window named “Mann–Whitney” pops up.
- Select “Overall Satisfaction 1” as the “First Sample.”
- Select “Overall Satisfaction 2” as the “Second Sample.”
- Click “OK.”
- The Mann–Whitney test results appear in the session window.
Model summary: The p-value of the test is lower than alpha level (0.05); so we reject the null hypothesis and conclude that there is a statistically significant difference between the overall satisfaction medians of the two customer types.
Comments are closed.