Concordance between power in logistic regression and slope of power curve given by equation 6.

Proportion of scenarios in which observed change in power agreed with predicted slope of chi-square power. The observed change in power is calculated as the sign of the difference of the median likelihood ratio statistic at γca = 0 and at γca = 1. 160 parameter values were considered, a subset of the parameter space described in table 1. 500 realizations at γca = 0 and at γca = 1 of each combination were undertaken to find the median likelihood ratio statistics.