MATH 203 Lecture Notes - Lecture 24: Wilcoxon Signed-Rank Test, Parametric Statistics, Statistical Power

• School: McGill University
• Department: Mathematics & Statistics (Sci)
• Course: MATH 203
• Professor: Michael Wallace

Document Summary

3. 4 comparing two dependent samples: within-subject data are dependent. Paired data: oten we may measure the same e. u. more than once. A parametric test assumes that the within-individual diference: xi= (yi1-yi2) are normally distributed, and test the null. H0: = 0 ( the true mean diference b/w the measurement sets is. But what if we"re not sure for the normality. We test a hypothesis about the , not : h0: = 0 (true median diference b/w the measurement sets is 0) Wilcoxon signed rank procedure: for paired pre:(y11, , yn1) and post:(y12, , yn2) Form xi= (yi1-yi2) for i= 1, , n. Drop every xi = 0, leaving a sample size of m. Assign ranks r1, , rm to s1, , sm ater soring into an ascending order. T+ 0 (the value has decreased) T-