Mann-Whitney's criterion is applied to assessment of distinctions on the level of expressiveness of any given sign for two incoherent or independent samples. These samples can differ on number of examinees. Mann-Whitney's criterion is especially reliable at the number of examinees which is not exceeding 20 people.

## It is required to you

- - 2 groups of examinees;
- - results of an experiment;
- - tables of critical values;
- - calculator.

## Instruction

1. Make an experiment and perform measurements in a scale of intervals or the relations. Samples have to be independent. In groups the number of examinees has to be more or equally we rub, or in the first it is more or equally 2, and in the second more than 5.

2. Write down results of measurements of both groups of examinees. Range data in both groups and present as one row. Place elements on extent of increase of sign.

3. Appropriate ranks to values of an integrated row. The value is less, the rank is less. At the same time the number of ranks has to be equal to number of results.

4. Divide a general series into two groups corresponding to the first and second sample. Find the total amount of ranks for each of them. Define big of the sums of ranks, corresponding to the first or second sample.

5. Determine the size of criterion of Mann-Whitney by a formula U=(n1*n2)+ (n+1)/2-R where instead of n1 specify number of elements in the first group, instead of n2 – number of elements in the second group, instead of n – number of elements in group with the largest sum of ranks, R – the sum of ranks, large in size.

6. For the chosen level of statistical significance determine critical values of criterion for the taken samples by tables of critical values. Draw a conclusion. If the calculated value of criterion is less, than in the table or is equal to it, then recognize existence of essential difference between level of the considered sign in the taken samples – the alternative hypothesis is confirmed, and zero it is rejected. If the calculated value of criterion is more tabular, then the null hypothesis is confirmed. The value of criterion is less, the reliability of distinctions is higher.