I've got the following data values:

POW |157 days after release | 379 days after release
1   | 2.46                  | 3.73
2   | 4.11                  | 5.46
3   | 3.93                  | 7.04
4   | 4.51                  | 4.73
5   | 4.96                  | 4.71
6   | 4.42                  | 6.19
7   | 1.02                  | 1.42
8   | 4.30                  | 8.70
9   | 7.56                  | 7.37
10  | 7.07                  | 8.46
11  | 8.00                  | 7.16

The values above are the test scores on brain function for Prisoners of War. The higher the test scores, the worse their brain is performing. I am trying to see whether the prisoners' brains perform worse 379 days after release than compared to 157 days after release.

Because the samples are performed on the same subjects, and the samples are assumed to be not normal and are less than 25 in size, I am supposed to use a wilcoxon signed rank test. The test is with a significance level of 0.05.

I realize how to rank and assign the differences, and how to get the value of 14 from the Wilcoxon table. I also realize that the negative rank sum is 9 and the positive rank sum is 57. What I do not understand is why, in the solution below, we use the negative rank sum (9) instead of the positive rank sum (57) in order to test against the value of 14 (we got from the Wilcoxon table). As well, why are we checking whether 9 is less than or equal to 14 (e.g., why don't we test whether 57 is greater than or equal to 14?)?

The solution to this question is apparently the following below:

enter image description here

  • $\begingroup$ What tables are you using? $\endgroup$ – Glen_b Oct 18 '16 at 3:22
  • $\begingroup$ I used this table: onlinepubs.trb.org/onlinepubs/nchrp/cd-22/v2appendixc_files/…. Where n = 11 and significance is = 0.05. $\endgroup$ – Kelsey Oct 18 '16 at 3:26
  • $\begingroup$ The information would be both in the book where it describes the test and normally in the margin or heading or footer of the table, but the heading and footer are cut off here. It's possible to work it out but I won't have time to do it for several hours $\endgroup$ – Glen_b Oct 18 '16 at 3:35
  • $\begingroup$ Okay, no worries, I appreciate the input, and I believe this enough for me to get by on for now. Thanks, Glen. $\endgroup$ – Kelsey Oct 18 '16 at 3:36
  • $\begingroup$ Well, that doesn't explain it, because your document defines it as the sum of the positive ranks onlinepubs.trb.org/onlinepubs/nchrp/cd-22/v2chapter6.html $\endgroup$ – Glen_b Oct 18 '16 at 3:45

You say "the" Wilcoxon table but there isn't a "the" for the Wilcoxon table -- there are several different tables to go with different (but equivalent) definitions of the test statistic.

Some tables are based on the smaller of the two statistics. That might be the case for you.

  • $\begingroup$ Ok, is there any way to know for certain which table should be used in my case, or have I not given enough information in order to do that? $\endgroup$ – Kelsey Oct 18 '16 at 3:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.