The main remaining challenge is to identify nuisance variables that would have a significant impact on the variance. This is called *analysis of variance* (or *ANOVA*, pronounced ``ay nova''), and methods that take this into account are called *ANOVA design*. Gender was an easy factor to imagine, but others may be more subtle, such as the amount of FPS games played among the subjects, or the time of day that the subjects participate. The topic is far too complex to cover here (see [154]), but the important intuition is that low-variance clusters must be discovered among the subjects, which serves as a basis for dividing them into blocks. This is closely related to the problem of *unsupervised clustering* (or *unsupervised learning*) because classes are being discovered without the use of a ``teacher'' who identifies them in advance. ANOVA is also considered as a generalization of the t-test to three or more variables.

Steven M LaValle 2020-01-06