Significant variables (there were no trend variables) from the Pearson r analysis
were entered into a hierarchical regression model in order of significance from the
strongest to the weakest; alliance training programs (yes) was entered into the first block
and number of employees into the second block of the regression model. Two different
models had produced significant F values, and the /-statistic for both models was
significant for the constant. The VIF values of the two models were all well below 10
and the tolerance statistics all well above .2, thus multicollinearity was not a problem.