Differences in median estimates for the three milestones and the two setback questions for each sample.

Positive values indicate the first listed sample making an earlier forecast or giving a higher probability respectively. The p-value section refers to the p-value in a non-parametric bootstrap comparison of the medians for each pair of samples. The critical values used in this analysis were Bonferroni corrected (0.05/24 = 0.002) because of the multiple hypotheses tested for in these data (including several hypotheses discussed in the Supplemental Material). Asterisk indicated statistical significance.