COVID-19 and SIMPSON'S PARADOX!

COVID-19: The number of deaths decreased, so we do better than before or better than ON or NY. --- SIMPSON'S PARADOX!

Today, we often hear that fewer new positive cases or deaths reported in a #province or state are a sign of better intervention or end of the #epidemic.

While the observation is promising & the inference may be correct, this finding may also be a misinterpretation of #epidemiological#modelling or the #Simpson's #Paradox.

--- A downward #trend appears in one or more subgroups of the #population, may disappear or be reversed when these groups are combined.

As long as the size of the disproportionately affected subpopulations is large or the underlying correlations such as deaths from #COVID-19 among residents of health care facilities are strong, or the new treatment protocol such as hydroxychloroquine works, the possibility of Simpson's Paradox must be considered when interpreting the results. --- Despite the development of the epidemic, a decrease in the number of deaths could've occurred.

In turn, the comparison of the number of deaths in 2 provinces or states is also not reliable.

Interpreting a trend, with or without statistical tests, requires standardizing the underlying associations as well as conservatism than looking at a single curve!