Nr. Sharpe et Ra. Roberts, THE RELATIONSHIP AMONG SUMS OF SQUARES, CORRELATION-COEFFICIENTS, ANDSUPPRESSION, The American statistician, 51(1), 1997, pp. 46-48
We examine the relationship between the extra sum of squares SSR(x(2)/
x(1)), the regression sum of squares SSR(x(2)), and the correlation co
efficients r(y1), r(y2), and r(12). From this we develop a necessary a
nd sufficient condition for suppression in terms of the correlation co
efficients, We use this to investigate the conditions under which supp
ression can occur algebraically and graphically. We believe that expre
ssing suppression in terms of correlation coefficients may help studen
ts and applied researchers to identify cases of suppression and to und
erstand when suppression can occur.