3 ms·

You know thats how study analysis in life science studies work don't you? Check out hazard ratio - https://en.wikipedia.org/wiki/Hazard_ratio https://en.wikipe

You know thats how study analysis in life science studies work don't you? Check out hazard ratio - https://en.wikipedia.org/wiki/Hazard_ratio https://en.wikipedia.org/wiki/Hazard_ratio - it covers the increase in likelyhood above base case. You seem to be implying it doesnt matter because the absolute number is low. Thats not how life science studies work. If this causes a 10k% increase in your hair turning orange, even if it affects a low number of people, its still a drastic increase in hazard ratio.

dahart 4y ago
The things you said are correct. However, it is also true, to parent’s point, that the error bars grow relatively larger the smaller the absolute sample size, and the smaller the measured effect is, right? This is also how life sciences work; noise in the data is more problematic the smaller the sample, and certainty is only achieved when the signal is far greater than the noise. (These researchers aren’t claiming certainty, they’re using this data to say more study is needed.) The abstract says “For overall mortality, with 74,193 participants and 61 deaths (mRNA:31; placebo:30), the relative risk (RR) for the two mRNA vaccines compared with placebo was 1.03 (95% CI=0.63-1.71).” A single person change in the death rate for either group here can change the summary outcome that people use as their talking point. Meanwhile, out of any other sample of seventy thousand people, it’s pretty likely that both numbers will change, probably by a larger delta than 1. The real problem here is drawing too hard of a conclusion like the GP did: “Danish researcher finds that one is more likely to die after taking the mRNA vaccines, even after accounting for the protective effects against covid.” Framing things this way is premature and misleading when you’re talking about 31 people out of 75k. The researchers did not come to that conclusion, they did not necessarily find that everyone is more likely to die with the vaccine than without, they found a small sample that perhaps suggests it might be possible, and they’re saying we should collect a larger sample to find out if it’s true.

dahart 4y ago

The things you said are correct. However, it is also true, to parent’s point, that the error bars grow relatively larger the smaller the absolute sample size, and the smaller the measured effect is, right? This is also how life sciences work; noise in the data is more problematic the smaller the sample, and certainty is only achieved when the signal is far greater than the noise. (These researchers aren’t claiming certainty, they’re using this data to say more study is needed.) The abstract says “For overall mortality, with 74,193 participants and 61 deaths (mRNA:31; placebo:30), the relative risk (RR) for the two mRNA vaccines compared with placebo was 1.03 (95% CI=0.63-1.71).” A single person change in the death rate for either group here can change the summary outcome that people use as their talking point. Meanwhile, out of any other sample of seventy thousand people, it’s pretty likely that both numbers will change, probably by a larger delta than 1. The real problem here is drawing too hard of a conclusion like the GP did: “Danish researcher finds that one is more likely to die after taking the mRNA vaccines, even after accounting for the protective effects against covid.” Framing things this way is premature and misleading when you’re talking about 31 people out of 75k. The researchers did not come to that conclusion, they did not necessarily find that everyone is more likely to die with the vaccine than without, they found a small sample that perhaps suggests it might be possible, and they’re saying we should collect a larger sample to find out if it’s true.