Calibrating the regression model on the benchmarking
data set, coef®cients of variables other than
PCLAY and CALCA had excessive standard deviations.
However, the signs of the coef®cients of all the
predictors in the model Eq. (1) are the same as those
calculated for the benchmarking model, and the con-
®dence limits (p0.05) of the coef®cients of the two
sets overlap (Fig. 3).