4

I'm looking at some microarray data. For the first time I've calculated a moderated T statistic from limma.

Is there any standard practice for where to cut off that value? For log2 fold change I usually cut off at +/- 1.5 and adjust accordingly; for adjusted p value I cut off at p<0.01 and adjust accordingly.

These moderated T statistics seem a little higher than log2 fold change and I'm not sure what is generally accepted as significant.

Thanks!

julianstanley
  • 401
  • 3
  • 9
  • 1
    If you're filtering by the (adjusted) p-value, why do you want to also filter by moderated T-statistic? The p-value basically comes from the moderated T-statistic, after all. – Devon Ryan Oct 19 '17 at 18:57
  • I would like to have values that are real but are also above a particular magnitude. The P value gives me an idea of whether the finding is real, but I also need a way to cut-off on magnitude. I'd usually do this by filtering l|og2Fold|>1.5, but I was hoping to have a similar moderated T-statistic cutoff. Or am I misinterpreting the moderated T statistic? – julianstanley Oct 19 '17 at 19:01

1 Answers1

3

You're misinterpreting the moderated T-statistic, it's basically the fold-change divided by its variance. The p-value comes directly from that, so if you filter by moderated fold-change you're just setting an unknown (unless you go through the trouble of figuring it out) p-value threshold. Instead, do exactly as you've been doing and set appropriate (adjusted) p-value and fold-change thresholds.

Devon Ryan
  • 19,602
  • 2
  • 29
  • 60