Second-class risk
Estimators and hypothesis testing

So far, we have considered a risk of error, i.e. a risk α of making an erroneous decision, which now needs to be explained in more detail. Vis-à-vis a hypothesis H₀ , for example the null hypothesis, α is the risk of rejecting the hypothesis H₀ when it is in fact correct. Typically, this is the supplier's risk. If we call H₀ the hypothesis that a manufacture is acceptable, this is the risk of making the decision to consider as unacceptable a batch that is in reality acceptable.

But we must also consider the other way of being wrong: this is the β risk of not rejecting the H₀ hypothesis when it is in fact incorrect. Typically, this is the customer's risk. To return to the previous example, it's the risk of taking the...