#fixed
17 messages · Page 1 of 1 (latest)
the picture is simply an ai breaking down the issue and their answer to it which i dont quite understand
i guess i dont get the whole "statistical power" side of things
What the AI is telling you is practically slop lol
i thought so
Don't rely on this stuff for understanding
It's only a good tool when you already know what's up.
also if it helps the 1000 was rejected 3 times 100 rejected 1 time and the 20 rejected 0 times
Check what happens across the board to try and understand what's going on. Simulate for small sample sizes (20 to 50) first. Then go larger (1000-2000).
Also. You have a misconception in your reasoning. More fluctuations doesn't mean more rejections.
ok but does it make sense for the rejections to increase?
also i just simulated 30 40 and 50 with rejections being 1 1 2 respectively
Yes. Look at the χ² statistic formula. That should itself tell you that you could have high fluctuations in count (with low sample sizes) but low rejections and contrastingly with large sample sizes even small deviations can be significant.
ok will do
You need to figure out what the right sample size is. One of the ways to do that using brute force is sweeping through sample sizes in your simulation and look for stable results.