ABSTRACT
Within the ‰eld of probability and statistical science, the reversal paradox is best known for categorical variables as Yule’s paradox or Simpson’s paradox. This is because George U. Yule noticed this phenomenon as early as 1903 (Yule 1903) when he referred to a paper published by Karl Pearson in 1899 (Pearson et al. 1899). The issue was later mentioned and made famous in a paper by Edward H. Simpson in 1951, discussing the way in which the relationship between two variables changed after a third variable was factored into a 2 × 2 contingency table (Simpson 1951). When such data are analysed by regression methods, the reversal paradox is more often known as Lord’s paradox, particularly within the behavioural sciences (Lord 1967, 1969), ever since Frederic M. Lord published his paper on this phenomenon with respect to the use of analysis of covariance (ANCOVA) in 1967 (Lord 1967). We discussed Lord’s paradox in Chapter 4. Within any generalised linear modelling framework, this phenomenon is more generally known in the statistical literature as the suppression effect, with the third variable termed a suppressor (Horst 1941; Cohen and Cohen 1983; Lewis and Escobar 1986; Lynn 2003; Friedman and Wall 2005).
