Are all qPCR experiments require Bonferroni corrections of p value to avoid hyperinflation of type I error which arises from multiple testing?
Honestly, whether to use a p-value correction or not is up to the analyst.
You are correct that as you conduct more hypothesis tests, the probability of getting at least one p-value below your nominal alpha value, just by chance, increases.
However, there are cases in which you may accept the probability of "false positives" in order to not miss likely "true positives". One way to think of this is how "conservative" you want to be with reporting potentially "false positives."
By the way, Bonferroni is a (excessively ?) conservative approach. You might look at other methods available for family-wise error rate (FWER) or false discovery rate (FDR).
Also, there is a consideration as to what constitutes a "family" of hypothesis tests where p-value correction might be appropriate.
And, finally, you are likely to encounter advisors or reviewers who insist that p-value correction is required in some circumstance. To me, it really depends on the purpose of the study and the goals of the analyst.
Alpha- or p-adjustment are needed in screening experiments that should identify one or a couple of candidates that are claimed (confirmed) to be regulated in the estimated direction.
Typically, such screening experiments are more explorative than confirmative. Using p-values to reject some null hypotheses is just not helpful or meaningful. You can use the p-value (adjusted or not - that does not matter!) as a proxy for a signal-to-noiss ratio, and to sort the list of genes according to this ratio. If you then pick some of the genes from the top of the list (possibly because a good signal-to-noise ratio seems attractive to you) and design subsequent experiments (knock-out/overexpression) to actually confirm their biological role in the process you are studying, then the initial p-value becomes irrelevant anyway.
Hello Asmaa Mahmoud Abdelmaksoud. I think the two 2005 Lancet articles on multiplicity by Schulz & Grimes are very thoughtful. You may find them helpful.
Article Schulz K, Grimes DMultiplicity in randomised trials I: endpo...
Article Multiplicity in randomised trials II: Subgroup and interim analyses
HTH.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA