• The conventional criterion for statistical significance is  that there is less than a 1 in 20 chance (5% probability, p < .05) that an observed  difference between two experimental groups could have occurred by chance alone even if there is really no difference.

• The more statistical comparisons that are made between two groups, the greater is the likelihood that one will yield  p < .05. With 20 comparisons, the likelihood that one comparison yield p< .05 approaches 100%.

• Statistically “positive” subgroups in overall negative trials should be viewed as unproven hypotheses unless the subgroup analyses were specified a priori  and corrected for the effect of multiple comparisons.