Hi Jon or another professional,
With n = 12, I recently demonstrated a Friedman test, measuring subjects' flexibility scores (treated as ordinal data) across four time measures. The omnibus test was significant, and asymptotic p values are below and ON THE LEFT (before Bonferroni corrections) showing the six pairwise comparisons taken from the default pairwise comparison matrix of the results section; to the right are asymptotic p values when the pairs were compared by conducting six separate Wilcoxon signed rank tests. Can you explain the differences in the p values for the two approaches?
pre-week3, p = .813 p = .317 (signed rank test)
pre-week6, p = .114 p = .026 (signed rank test)
pre-week8, p < .001 p = .002 (signed rank test)
week3-week6, p = .179 p = .040 (signed rank test)
week3-week8, p = < .001 p = .002 (signed rank test)
week6-week8, p = .009 p = .003 (signed rank test)
I might expect differences if the p values were compared to exact p values produced by using the T-plus distribution table but not with both results being asymptotic p values. Thanks in advance,
Cliff