I am unsure whether I have to correct for multiple testing in my case. I have symptom severity ratings for several (9) symptoms across 3 time points (pre, during, post). I want to test whether "during" predicts "post" over and above "pre". So I´m comparing regression models. I do that for each symptom separately. Since I have several separate regressions that share neither IV nor DV, do I have to correct for anything? I hope you can help me!
Asked
Active
Viewed 44 times
0
-
1Multiple testing when the null hypotheses are true increases the probability of (at least once) erroneously rejecting a null hypotheses. If spuriously producing "statistically significant" results is a concern then you should adjust for this. – Henry Nov 09 '16 at 12:02
-
Related: https://xkcd.com/882/ and http://stats.stackexchange.com/a/200640/2958 – Henry Nov 09 '16 at 12:03
1 Answers
0
Partially answered in comments:
Multiple testing when the null hypotheses are true increases the probability of (at least once) erroneously rejecting a null hypotheses. If spuriously producing "statistically significant" results is a concern then you should adjust for this.
– Henry
kjetil b halvorsen
- 63,378
- 26
- 142
- 467