SAMPLE SIZE: The function of sample is to fairly represents the population. In so doing, the sample must not be too large so as to be inefficient, nor be too small so as to become bias. Thus, the right size is "minimum sample size."

FINITE POPULATION: If the population size is known, then the Yamane method is a common practice. N = population size, and e = error level. The sample size is given as:

[1]   n = N / (1 = N(e2)

However, this method may not be feasible for cellular research where dealing with tissue count or cell count would lead to non-finite population, i.e. some time the tissue available is limited or cell death or growth may be unpredictable. Non-finite method may be appropriate.

NON-FINITE POPULATION: In non-finite approach, the sample size may be attained by:

[2]   n = (Z2σ2) / SE2

where Z = critical value for a specific confidence interval; σ = estimated standard deviation; SE = standard error obtain by SE = σ / sqrt(nt) and nt is the test sample size. Under this method, it is necessary to take a test sample first and obtain the necessary descriptive and inferential statistics before using equation [2] above.

In Alan Gresti, n = 30 is a common practice in order to obtain distribution information and proceed to minimum sample size calculation. However, In Anderson-Darling, the minimum size for test sample may be as low as n > 5. ultimately, even when the "adequate" sample size is obtained under [1] or [2], one would still do simulation to get a result of repeated measurements in the later testing.

Some references and links are attached. A book by Montgomery is a good reference, see link below.  I trust this has been helpful. Cheers.

https://mathdept.iut.ac.ir/sites/mathdept.iut.ac.ir/files/AGRESTI.PDF

http://www.stat.ufl.edu/~aa/

http://guilanstat.ir/wp-content/uploads/Douglas-C.-Montgomery-Design-and-Analysis-of-Experiments-Wiley-2012.pdf