My data does not follow a normal distribution, but Mann Whitney U test gives me a p-value of 0.935?

[Aaron4174]

As the question indicates, I conducted a normality test on Minitab, and the p-value I received was 0.018, which is less than the significance level, therefore, it is not normally distributed. Therefore, I had to conduct a non-parametric Mann Whitney U test, and the p-value is 0.935, which indicates the dataset is not statistically significant?

What does this mean?

 

[GretaGarbo]

It means that it was not statistically significant.

What were the means and standard deviations when you did a t-test?

 

[hlsmith]

So you did a normality test and rejected the null of normally distributed. Normality tests are known to fail if sample size is large. What was your sample size? It is usually best to plot data in histograms and q-q plots to examine normality. If normality is close, which a 0.02 could be, and sample size large enough, a ttest may still be used. Though you will need to test homogeneity of variance.

The above U test pvalue conveys something about the ranking of medians, you just need to look it up online, but that pvalue is pretty darn high I will note. I will also note that if your sample size is low and data non-normal, there is an exact wilcoson test and approximate version of the Utest that is Monte Carlo. Tell us more about your data and provide graphical representations please.