Hi all
Scenario, there is 40 participants, divided into two, 20 participants in each group.
If z score has no outliers (+-3.29) but there is clearly one in boxplot and Q plots and the test of normality from Shapiro is met on one score but not met on one score. Now, the question on the paper is to use z score and normality test only. Since Shapiro is not normal, I used a non parametric test to use which is Manwhitney U test.
In a real case scenario, you would correct the outlier in boxplot and continue with parametric (independent t test). I could ignore the unmet normality test since I have 20 participants that can tolerate normality test.
Any know much about this?
From what you've said, I assume that your plan was to compare the two groups using an unpaired (independent groups) t-test. Is that right? If so, what is the DV, and what do you know about its (population) distribution from sources external to your study? I have in mind the advice given in this article by Bland & Altman (2009):
Article Practice Statistics Notes Analysis of continuous data from s...
HTH.
The normality tests are very often misinterpreted. Note that for both S-W AND
K-S large p-values favor the normal. The easiest place to look that up is WIKIPEDIA, look at either the particular test or goodness of fit tests. Hope this helps. Best wishes David Booth
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size