2. Identify outliers : Standard deviation test

Beyond sampling issues, the question of so-called outliers raises the problem of "abnormal" sampled values. If, for example, we are interested in the sum of a throw of five dice, obtaining a sum equal to "30" (five "6s") in a sample of ten throws raises questions, because the probability of this event in a small sample is practically zero. If you use this value to calculate estimators of the mean and standard deviation of the parent population, you will obviously find estimators that are quite far from reality. In this case, it's best to leave this value out of the calculations. This example shows how important it is to be able to detect this type of value.

There are many tests to detect them, the simplest of which is the so-called "normalized deviation" (ND) test. Any series of data can be normalized, i.e. "transformed" into a series of averages.