ABSTRACT
The pooled table apparently suggests a positive association between the row and column variables. However, the two subtables both suggest a negative association, indicating that the positive association in the pooled table should be interpreted as a spurious association. The spurious association is created by the lurking variable, nationality. Simpson’s paradox creates considerable problems with statistical interpretation, and all statisticians should remain continuously aware of this paradox and its ramifications. The chapter discusses a cross-classification of recovery against treatment, that there was an apparent positive association between treatment and recovery. A fuller analysis of these data may, however, be performed by closer consideration of the association between treatment and nationality, and between recovery and nationality.
