3 Datasets Hypothesis testing

#1
n = 72 subjects were randomised to 3 treatment groups: A, B, C, so that n = 26 for Group A, n = 29 for Group B, n = 17 for Group C. Measurements were taken before and after treatment. The differences were calculated between the before & after-treatment observations for the 3 treatment groups: A_diff, B_diff, C_diff. Shapiro test p-value for A_diff = 0.2567, Shapiro test p-value for B_diff = 0.007945, Shapiro test p-value for C_diff = 0.5156. What hypothesis test would I use? One-way ANOVA? Kruskal-Wallis Test? Suggestions for post-hoc tests?
 
Last edited:

Karabiner

TS Contributor
#2
The total sample size is large enough, so a oneway ANOVA is robust against non-normality
within groups. Homogeneity of variances is important. Post-hoc tests there are many. If
variances are homogenous, maybe Tukey.

With kind regards

Karabiner
 
#3
The total sample size is large enough, so a oneway ANOVA is robust against non-normality
within groups. Homogeneity of variances is important. Post-hoc tests there are many. If
variances are homogenous, maybe Tukey.

With kind regards

Karabiner
Let's say now, that Group A is a control group. Would Dunnett's post-hoc test work? Dunnett's Test is ideal for a balanced study design, so is Tukey better in this context?