CORRELATES OF REACTIVITY - PROBLEMS WITH REGRESSION-BASED METHODS

Computer simulations were used to compare change score and regression methods for identifying relationships between a third variable and rea ctivity. When the third variable was not correlated with baseline, the two methods were equally effective for detecting relationships betwee n the third variable and change. However when there was a correlation between the third variable and baseline, regression measures of change yielded a high rate of Type one errors. Regression methods also had l ess power for detecting a relationship between the third variable and change when the relationship was in the direction opposite to the sign of the correlation with baseline. The implication of these findings f or two areas of research are discussed: the relationship of fitness to heart rate reactivity and the relationship of hypertension to blood p ressure reactivity.