Timeline for T-test or Wilcoxon for small sample paired data
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 18, 2020 at 21:37 | comment | added | Christian Hennig | Does that mean you ran 400 Shapiro tests and looked at 400 plots? In any case, note that if you run 400 tests, there will be random significances. Only Bonferroni-corrected p-values (i.e., significant at 0.05/400) give a clear indication that a specific metabolite is different in such a situation. Given that with $n=5$ the power of any test is very weak, I don't see that you can get a single p-value of this kind. So chances are significance testing will not address this issue appropriately. You'd need more data. | |
| Aug 18, 2020 at 20:13 | answer | added | Eron Raines | timeline score: 1 | |
| Aug 18, 2020 at 19:58 | history | edited | BruceET | CC BY-SA 4.0 |
edited title
|
| Aug 18, 2020 at 19:51 | history | edited | BruceET | CC BY-SA 4.0 |
added 3 characters in body; edited title
|
| Aug 18, 2020 at 19:28 | answer | added | BruceET | timeline score: 1 | |
| Aug 18, 2020 at 15:59 | comment | added | Mee | I have around 400 metabolites for 5 paired samples (before and after a treatment) and want to know which metabolites are different before and after treatment. I know the data es not normal because of the plots and the shapiro test showed the data es not normal. I was thinking maybe about transforming the data... | |
| Aug 18, 2020 at 12:53 | comment | added | Christian Hennig | You are not stating what exactly you want to know. With 5 pairs the power of any test will be very low, so it may not make sense to run any test. The paired t-test could give some indication in case the data are not too non-normal. How do you know your data "do not follow a normal distribution"? What indication against normality do you have? And why do you need a test? | |
| Aug 18, 2020 at 12:01 | history | asked | Mee | CC BY-SA 4.0 |