Calculating t-test p-value for groups containing only 2 samples

Question

[yuyu]

Is there any meaning to calculate the p-value for groups of only 2 samples? I mean......I suppose that we can but is it right to do that with so few samples?

Answer 1

[Devon Ryan]

This is more appropriate for Cross Validated. Having said that, no, there isn't much point to running a T-test with only two samples per-group. You'll get a p-value, but you'd be well advised to not read much into it unless you happen to know something a priori about the within-group variance.

[yuyu]

Then how about calculating the p-value between two samples in qPCR, each having 2 replicates?

[Devon Ryan]

I assume that the "replicates" are the normal qPCR technical replicates. Again, you can get a p-value, but it's hard to read much into that. A small number of samples will always result in high degrees of uncertainty.

[yuyu]

Yes my replicates are technical replicates. I agree with your opinion but the lab where I currently belong to wanted me to calculate this p-value and see if the difference between two samples are significant........I told them that the value will not mean much but they even said that many other publications do the same thing.... Anyway, thank you very much for the answer!

[Devon Ryan]

I'm sure they'd jump off a bridge if others were doing that too.