Given three variables, x, y, z , univariate selection on x , usually reduces the correlation r y z . However, in some cases, when r x y or r x z assumes extreme values, r y z may increase. This reversal is explained by comparison to the effects of suppressor variables and predictors of predictability.
No takes yet. Share an insight, caveat, or question.
Joseph Levin (1972) studied this question.
Synapse has enriched one closely related paper. Consider it for comparative context: