Similarity Tests are used to determine if you can generalize that two entire populations are truly different based on two samples:
- If both samples have 10-30 values and are normally distributed: use the T-Test
- If at least one sample has more than 30 values, and both are normally distributed: use the Comparison of Two Means
- If you are unsure whether both samples are normally distributed: use the Mann-Whitney U-Test
- If the data is in percentages (e.g. frequency tables), and the total of each column is equal to 100%, you can use the simple Index of Dissimilarity