**30**wrote:

When working with qPCR data, the general route to take is to obtain ddCT values by normalising to an internal reference gene followed by subtracting from the mean of the control values. This is followed by raising 2^-(ddCT) to obtain the final gene expression value.

We generally work with pooled samples of 2 experiments where each experiment consists of 4 biological replicates being tested for the control and treated condition.

Browsing the web and literature, I see no consensus on whether to use parametric tests or non parametric tests for testing for (the difference in whatever chosen statistic) significance. I see that online discussions in general state- if data is normally distributed across all treatments, then use a parametric test and vice versa.

But is this the right way to go about it? Especially given that our sample size is never greater than 8, should I at all be considering parametric tests?

I find that testing the data for normality yields contrasting results when I examine the same biological specimen (given the same treatments) in independently conducted experiments. Should I always err on the side of caution or use non-parametric tests?

For example, if I were to test for the population size of *Drosophila* in 10 vials, I would know that irrespective of whether I could actually obtain a normal distribution in my sampling, biological populations in general follow a normal distribution and therefore I would always apply a parametric test if comparing with another population!

The data you have in hand (especially if it's n=8) should not necessarily look like normal distribution but it should be sampled from a normal distribution. So if you'll take a lot qPCR results will it look like normal distribution or not? (I honestly don't know but I suspect that the answer is yes, a transformation at least). The other issue is with n=8, t-test should only be applied for n>=20 (or 23, depends who you ask) so I would go with a non-parametric test in your case.

8.5k