Post by konrad on May 15th, 2013 at 4:16pm
Dear matias,
It depends on what you want to report. If you want a list for which you are confident (at a certain significance level) that there are no false positives in the list, you need to do a family-wise correction, e.g. Holm-Bonferroni. We (and others in similar contexts) have argued that this is typically not what one wants when screening multiple genes - instead, we suggest reporting the uncorrected list as is along with the q-values (see Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to
), as we did in the MEME paper (tables S4-S19). The interpretation is then that the list is expected to contain false positives, and each individual q-value can be interpreted (roughly; see Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to and Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to for a detailed description) as the probability that that particular result is a false positive. The sum of the q-values gives the expected number of false positives in the list.
So we advocate a philosophy under which we are willing to report lists that almost certainly contain false positive results, along with numbers from which we can estimate how many false positives there are likely to be, but not which ones they are. Of course, it is then important to make sure this is clear when reporting the list.
Hope this helps,
Konrad
It depends on what you want to report. If you want a list for which you are confident (at a certain significance level) that there are no false positives in the list, you need to do a family-wise correction, e.g. Holm-Bonferroni. We (and others in similar contexts) have argued that this is typically not what one wants when screening multiple genes - instead, we suggest reporting the uncorrected list as is along with the q-values (see Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to
), as we did in the MEME paper (tables S4-S19). The interpretation is then that the list is expected to contain false positives, and each individual q-value can be interpreted (roughly; see Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to and Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to for a detailed description) as the probability that that particular result is a false positive. The sum of the q-values gives the expected number of false positives in the list.So we advocate a philosophy under which we are willing to report lists that almost certainly contain false positive results, along with numbers from which we can estimate how many false positives there are likely to be, but not which ones they are. Of course, it is then important to make sure this is clear when reporting the list.
Hope this helps,
Konrad