I have the results of a questionary with 2 subjects and each of them includes 4 sub-questions in the form of Likert items. The questions are about the impact of a situation on people's lives and the options are negative, little negative, neutral, little positive, and positive. My main question is how to interpret the results to understand the direction of impact. I encoded them (negative=-2, little negative=-1, neutral=0, little positive=1, and positive=2) and used one-sample Wilcoxon signed-rank test and put the difference equal to 0 in python. I appreciate any help
Comment. I agree with @NickCox that some sort of graphical display may be the best you can do. If I understand correctly, you have one four-question questionnaire from each of two people. Even if the two subjects have completely non-overlapping opinions, I know of no appropriate test that would find a statistically significant difference
A paired Wilcoxon signed-rank test will not be able to give a P-value below 5%.
Consider the following fictitious data, with a paired Wilcoxon test in R, which shows no significant difference:
a = c(-1, -1, 0, 0) b = c( 1, 1, 2, 2) wilcox.test(a, b, pair=T) Wilcoxon signed rank test with continuity correction data: a and b V = 0, p-value = 0.07186 alternative hypothesis: true location shift is not equal to 0 Warning message: In wilcox.test.default(a, b, pair = T) : cannot compute exact p-value with ties