If you only have 10 people and important assumptions are violated, such as normality, then you may consider the Wilcoxon signed-rank test. This is quite a small sample, so power would especially be an issue anyway for a parametric analysis, such as the standard t-test.

Thank you for your answer. I have 3 sets of data that I want to compare: pre, post and follow up scores. Wilcoxon signed-rank test compares two sets of data I guess. So Friedman Test seems to be a good choice, or does it? Cory Cobb

The choice of an appropriate statistical test depends on the situation. Generally, you can use either a parametric test or a non-parametric test.

However, if you opt for parametric tests, the conditions of their validity are numerous and are only tracks which can change according to the type of data and specific fields of application. It is strongly recommended to refer to the recommendations specific to your fields. These conditions are as follows:

1) The measurements are independent

2) The population that generated the sample follows a normal distribution (assume or verify)

3) Samples have equal variances

4) The residues follow a normal distribution (suppose or verify)

5) At least 20 individuals per sample, or normality of the populations of each assumed or verified sample

6) At least 20 individuals in the device, or normality of the residuals assumed or verified

7) Each variable follows a normal distribution

8) At least 20 individuals in the sample (recommended)

9) No theoretical workforce below 5 in the boxes of the table

10) Differences between series follow normal distributions.

Ekin, the Friedman test seems appropriate - as long as your outcome is either ordinal or continuous....

Cory

Thank you for you answers Cory Cobb and Ben Boubakary

You are welcome dear Ekin D. Kozak