If you compared multiple groups and reported confidence intervals do you have adjust the confidence inerval for multiple comaprisons just as we do for P-values? If yes, How?
Thanks
Are you going to compare them by one factor or more? you can use the ANOVA test and highlight the existence or no difference and the underlying p_value.
if you talk about the multiple test, we usualy preced to use the Bonferroni correction for p-value and also for CI( see the attached fil)
Waiting for further clarification.
Yes, because (1-α)-CIs are just those intervals for which a test would not be significant at α. So if the test is adjusted for multiple testing, CI must be adjusted accordingly.
Some software do give you the adjusted CIs. For instance, TukeyHSD in R does so.
If you want to do an adjustment to hold the FWER manually, a simple (but somewhat conservateve) solution is based on the Bonferroni correction. If you have k comparisons, each individual test is performed on α/k to hold a FWER of α. Thus, to get (Bonferroni-)adjusted (1-α)-CIs you can calculate (1-α/k)-CIs.
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling