Timeline for Create simulation to see whether Chi Squared is suitable
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 2, 2019 at 21:21 | comment | added | Pere | About uniform distribution of p-values: P-values are computed by evaluating the inverse distribution function of a chi squared distribution on the the statistics. If the statistics is actually distributed according to that chi squared distribution, it will be uniformly distributed between 0 and 1. For example, 5% of all values of the test statistic will be under the 5% p-value threshold - and the same is true for any other percentage. | |
| Mar 2, 2019 at 20:42 | comment | added | baxx | example with different probabilities : vpaste.net/torSn , which still seems to work out alright even though ~40% of the cells have what could be considered a low count | |
| Mar 2, 2019 at 20:40 | comment | added | baxx |
thanks, I'm a little unsure how to interpret the part about the uniform distribution. Because the p-values are roughly uniform (I've plotted them from your code), we would reject ~5% if we were to reject values <0.05 (as in table(res<0.05)/s). If the test wasn't working so well would we expect to see this value increase / decrease? (I've tried using different probabilities, and it seems pretty robust)
|
|
| Mar 2, 2019 at 18:41 | comment | added | Pere | You have here one example code. I tried to copy your data but it doesn't seem to work. I think there is a problem with parentheses count and - maybe - with a too long line. | |
| Mar 2, 2019 at 18:34 | history | edited | Pere | CC BY-SA 4.0 |
expansion
|
| Mar 2, 2019 at 18:20 | comment | added | baxx | how would one go about writing a simulation for this though? I've added some example data to the OP | |
| Mar 2, 2019 at 18:15 | history | answered | Pere | CC BY-SA 4.0 |