Hacker News Reader

Comment by moralsupply | Hacker News Reader

> The trial involved 30,000 people in the US with half being given two doses of the vaccine, four weeks apart. The rest had dummy injections.

> The analysis was based on the first 95 to develop Covid-19 symptoms.

> Only five of the Covid cases were in people given the vaccine, 90 were in those given the dummy treatment. The company says the vaccine is protecting 94.5% of people.

Aren't those numbers way too small to make any statistically significant claims?

View on HN

Not at all. That is why the N on these trials is so huge. 5/15000 vs. 90/15000 is going to be statistically significant anyway you slice it. It’s a 45-fold difference. You can approximate it yourself with a t test.

A Fisher test you mean. T is for continuous variables.

Haha indeed

The difference between 5 and 90 is highly significantly different from equality.

And same as with the BioNTech vaccine, the testing continues until they have 160+ covid cases in one of the legs.

No. The full trial was on 30,000 people, every phase 3 trial picks a certain number of infections to stop at in order to draw results. Of the control group, 90 something got COVID, while only 5 of the vaccinated group got it in the same timespan. This is considered a big enough difference to call the vaccine effective (so far).

It’s not. I can’t remember all the details but if you read the moderna cove study 95 people after a month or two gives a very high statistical confidence.

The 94.5% value certainly seems like it has too many significant figures. If just 1 more person had happened to contract covid in the vaccinated group, the success rate would be reported as quite different.

Reply on news.ycombinator.com