How many samples are needed for a T-test Reading scientific biological papers I can see that comparing two conditions experiments are often repeated 4-6 times when a T-test is applied (for example cell survival is compared between control and drug-treated cells).
Some papers have as low as 3 replicates, the good ones might go for 7-10 replicates.
What should be a gold standard for a T-test?
 A: There is no particular sample size $n$ that is always best. For each
study you need to take a number of issues into account, and that
process can lead to the required sample size in a given situation.
Let's assume that you are testing at the 5% level, and that you
want the probability (called 'power') of detecting something of interest to be
high---perhaps 80%, 90% or 95%.
Then you need to decide the
amount of difference $\Delta$ between means of control and treatment groups
you want to detect. Also, if your observations have small variability, then
it will be easier to detect important effects than if variability is
large, so you need an educated guess for the standard deviation $\sigma$
of the observations. You will seldom be absolutely sure about $\Delta$ and $\sigma,$
so some guesswork based on previous experiments of the same kind is often
helpful.
With a choice of desired power and a good guess about the ratio of $\Delta$ to $\sigma,$
you are ready to use formulas or computer procedures to find what sample
size $n$ for control and treatment group you should use.
Suppose your significance level is 5% for a two-sided test, desired power is 90%, and that you
want to detect a difference $\Delta$ that is about $1.5\sigma.$
Then a sample size computation gives $n = 11.$
Here is output from Minitab's 'power and sample size' procedure for two-sample
t tests.  Notice that I input $\Delta = 3$ and $\sigma = 2,$ but any choice
with $\Delta/\sigma = 1.5.$ would have given the same sample sizes.
Power and Sample Size 

2-Sample t Test

Testing mean 1 = mean 2 (versus ≠)
Calculating power for mean 1 = mean 2 + difference
α = 0.05  Assumed standard deviation = 2

            Sample  Target
Difference    Size   Power  Actual Power
         3      10    0.85      0.886970
         3      11    0.90      0.916899
         3      13    0.95      0.956112

The sample size is for each group.


Most statistical software programs have power and sample size procedures
for a number of different kinds of tests that are in common use. (I happen
to like the output format provided by Minitab.) There are also Internet
sites that have relevant calculators of varying degrees of accuracy and
ease of use.
A: Probably BruceET will kill me... but in my work environment the idea is that for a TTest  more than 6 sampkes are needed to be able to apply a T test.
The idea is that assuming a notmal distribution, 7 samples is a decent number to represent the distribution were 2 -3 samples can no represent it .
I know this is not correct and all the Pro here will disagree and they could write numbers and formulas. Sadly there is another reality on where many people who uses these tools but their level of statistics is basic.
If what i said is wrong, many people are doing these wrong. If there is something true here, maybe someone can expand my answer.
