home *** CD-ROM | disk | FTP | other *** search
- MANN-WHITNEY U-TEST -OR- TWO SAMPLE RANK TEST
-
- AIM: This is a non-parametric test. Such tests make no assumptions about
- the normality of the population distribution and do not require that the
- measurements be very precise, provided only that they may be arranged
- meaningfully in a rank order. It is applied here to two sample means, and
- the program calculates U and U`. To test for significance, a table,
- "Critical Values of U in the Two Sample Rank Test", is consulted.
-
- EXAMPLE: In an experiment, five control (untreated) and six experimental
- (drug treated) rats were studied. Blood glucose was measured in each rat.
-
- BLOOD GLUCOSE, mg/ml FORMAT FOR DATA ENTRY FROM FILE
-
- CONTROLS DRUG TREATED Controls <--- title for first group
- 2.15 2.02 5 <--- N for first group
- 1.92 1.71 2.15 <--- data point 1
- 1.78 2.04 1.92 .
- 2.04 1.50 1.78 .
- 2.22 1.83 2.04 .
- 1.69 2.22 .
- Drug Treated <--- title for second group
- 6 <--- N for second group
- 2.02 <--- data point 1
- 1.71 .
- 2.04 .
- 1.50 .
- 1.83 .
- 1.69 .
-
- Note that the raw data is rearranged in ranked form, and both the raw data
- and the results are printed out. Note also that 5.5 + 24.5 = 30, and that
- 6 X 5 = 30. The smaller U (5.5) is compared with the critical value in the
- table mentioned above. In this case, the critical value for N = 5 and N` =
- 6 in the table is 5, at the 5% level. Since 5.5 is not less than 5, the
- drug was not effective at the 5% level.
-
- REFERENCE: Goldstein, A. 1964 Biostatistics: An Introductory Text.
- Macmillan Company, New York. Pages 55-57 and Table 15.
-
- BASIC VERSION: Dr. Stanley Kaplan Sep 83
- PASCAL VERSION: Dr. Stanley Kaplan Oct 85
- UPDATES: Oct 86 (Version 2.00)
- Feb 87 (Version 2.10)
- Feb 88 (MS-DOS Version 2.2)