Quantification of individual sensitivities to ethanol in selective breeding experiments: difference scores versus...

Nagoshi et al. (Alcohol Clin Exp Res 10:343-349, 1986) have argued recently that the use of postdrug minus predrug difference scores to assess individual variability in sensitivity or tolerance to ethanol is not to be recommended. Unreliability of variables can result in biased estimates of response to treatment. Instead of difference scores, they recommend using residual scores from the regression of posttreatment on pretreatment measures. We were interested in the consequences of employing difference scores versus regression residuals for a related, but different purpose: the choice of individual animals to be mated for a selective breeding study. The availability of data from such a selective breeding experiment ongoing in the laboratory of one of us (JCC) allowed us to compare directly the effects of use of change and residual scores as criteria for selection. The selection is for maximal hypothermic sensitivity of mice to an acute dose of ethanol. In our sample, difference scores were highly correlated with regression residuals. In practical terms, almost 75% of the individuals chosen for mating by the change score method would also have been chosen by the residual score method. We conclude that for selective breeding experiments, the use of difference scores does not necessarily introduce sampling bias, is unlikely to affect the ultimate outcome of selection, and may be a more meaningful variable for the physiological process studied.