To test whether predictor loadings for logistic regression analyses of subjects choices were significantly different from zero, bootstrap confidence intervals on the population means μ were evaluated by generating a set of N=5000 resampled datasets by sampling subjects with replacement. P values for predictor loading significantly different from zero were calculated as:31P=2min(MN,1−MN)WhereM is the number of resampled datasets for which μ > 0.
