Precious Roy wrote:
A responsible scientist would not publish findings with a huge uncertainty interval generated by comparing @100 subjects with @4900.
What they did was take a population of about 5000 people. About 20% of those were at least occasional joggers. Of the joggers about 10% (2% of the total population) were "more than occasional" by their criteria. They compared sub-populations from this plausibly unbiased sample and generated statistics including uncertainty intervals consistent with the size of the sub-populations. Per those calculations they have drawn conclusions with specific confidence levels. One of their conclusions is that strenuous jogging by their criteria is correlated with a much higher hazard ratio than light jogging by their criteria and the difference is statistically significant at the 95% confidence level.
They may have done something unreasonable in this process, but comparing different sized sub-populations isn't it.