Fig.1(B) leads to a smaller variance compared to independent situation Fig.1(A) with the same number of samples. Since larger sample size leads to smaller variance, it is as if the “effective sample size” is increased in Fig.1(B). The trend in Fig.1(C) is the opposite: the “effective sample size” is actually reduced. Take an extreme example of monozygotic twins: since monozygotic twins have identical genotypes, a pair of twins provide the same genetic information as one twin, and the two points within a circle in Fig.1(C) is equivalent to one point. In other words, the effective sample size is only half of the apparent sample size. These concepts have already been understood in the study of clustered/clumped data and are associated with phrases like “variance inflation” and “overdispersion”.
