LAUSR

In Hunt et al., we used the Anderson–Darling (AD) two-sample test to evaluate the probability that the observed linear rank distributions for female and male principal investigators are from the same parent sample. We determined the nominal AD p-value, ρ AD(x, y), where x and y are the female and male linear ranks, and estimated the uncertainty by calculating the mean and 68% confidence intervals from the AD p-value bootstrapped distributions, ρADboot . Since the p-value is a measure of probability, and the distribution of p-values is uniform, this method is not appropriate. Here we calculate ρADboot as the proportion of bootstrapped AD test statistics that are greater than or equal to the nominal AD test statistic. We find that the nominal and bootstrapped AD p-values are similar and therefore we do not recommend any changes to the conclusions in Hunt et al.