Oh boy. Someone doesn't see any difference in 14.9 and 12.7 but hails the difference in 12.7 and 10.4.
Wow. Thats some gymnastics that you literally said would need to be reviewed further.
Since I am lecturer, I give free tutorial.
We have a process similar to tossing a coin. With a fair coin it is heads half the time and tails half the time.
In this case, the process is whether someone with covid ends up in a hospital. We see from the placebo sample it is roughly 14%.
The variance of adding 700 observations like this is:
700(p)(1-p)
where p is probability of the event happening...in the case of a tossing a fair coin p is 0.5...in the case of a person going to the hospital it is 0.14
a standard deviation is the square root of the variance
using the formula with a little help from my best friend excel we get a standard deviation of slightly over 9
expected value of how many of 700 will go to the hospital is .14 times 700 or 98
these are the key numbers: expected value of 98 and standard deviation of 9
this means if you give several groups of 700 the placebo, about one third of the time you will get a number less than 89 (mean minus one standard deviation or higher than 107 (mean plus one standard deviation).
So for ivermectin to outperform the placebo by nine is to be very close to outperforming it by one standard deviation...which in random draws involving groups of 700 all taking just the placebo would happen about one third of the time
So now we move on to fluvoxamine, which outperformed the placebo by 32. That's slightly more than 3 standard deviations.
This is why the people who did this trial are convinced of fluvoxamine and not of ivermectin.
One outperformed the placebo by 3 standard deviations (which has less than 1% chance of happening if it were worthless)
The other outperformed the placebo by 1 standard deviation (which has about a 33% chance of happening if it were worthless)
Anyhow I hope I got all that right.
