Hacker News Reader

> I'm not sure how 21% lower is considered "not statistically significant", in trying to suppress the spread, ANYTHING > 0% is helpful. Full stop.

Statistical significance has a specific meaning in the context of hypothesis testing. It is a measure of likelihood that the observed result occurred due to a real difference between groups (rather than random chance).

View on HN

It seems that they are adding up the margin of errors for 82/1461 and 87/1461, (schools responded divided by schools surveyed), giving a total margin of error of ~20% for these optional vs. mandatory masked student statistics. This is a problem with using surveys with a low response rate.

In their own words in that section, by the incident rate ratio it is statistically significant, even after having been adjusted for county level 7 day incidence.

You can try and figure it out on page 4 of the cdc report, it does not appear to be a null hypothesis test.

https://www.cdc.gov/mmwr/volumes/70/wr/pdfs/mm7021e1-H.pdf

Reply on news.ycombinator.com