Hello,
I am running an analysis on dietary information, comparing nutrient intake between two independent groups. There are approx 25 nutrients being assessed, so I need to run that number of independent t tests. I am concerned running this large a number of t tests will significantly increase the chance of seeing a false positive. Is a Bonferoni correction the best method to use here (i.e. significance level set at 0.05/25 = 0.002), or is there a better method to utilise? Is it definitely necessary to adjust the p value?
Thanks in advance,
Rob
Hi Robert Browne
You can adjust for False Discovery Rate (FDR) which is less conservative than the Bonferoni method. For details see: https://www.statisticshowto.datasciencecentral.com/false-discovery-rate/
Some researchers also argue not to adjust at all. For example, see: Article No Adjustments Are Needed for Multiple Comparisons
Thanks so much Usman, the Benjamini-Hochberg method looks exactly what we need.
- Badges
- Science topic
- Similar topics
- Correction