I run a Wilcoxon test for the folowing data: pretest: 63 75 75 74 65 63 and post test: 70 48 65 60 68 57
I obtained a W of 17 for a p level of 0.219. I used SPSS for the same data, I have a p level of 0.173 for a Z value of 1.362. I have 2 negative ranks and 4 positive. Why the results are so different? When a do a t test the value of t is the same for the three software.
Moreover, the matched rank biserial correlation is suppose to be the difference between the number of + ranks/total number of ranks minus the number of - ranks / total number of ranks. The result of 4/6 minus 2/6 is supposed to give 0.333. In the ouptut, I have 0.619.
Thanks for your future comments
Dear Gilles Dupuis, The first result is obtained from the exact null distribution of the Wilcoxon statistic for two paired samples. This is the correct result for the sample size used.
The second uses (incorrectly for n = 6) the normal approximation, Z = (W - EW) / sqrt (var (W)), and finally the "approximate" p-value is calculated for this Z (two-tailed). Check the exact calculation option in SPSS.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running