If your sample size is 2 some program will probably take it. That's not how you decide what size sample to take. See https://www.google.com/search?ei=ykXGXrP-DszOtQa8goioBw&q=what+size+sample+should+I+take+based+on+statistical+methods&oq=what+size+sample+should+I+take+based+on+statistical+methods&gs_lcp=CgZwc3ktYWIQAzIFCCEQqwIyBQghEKsCOgQIABBHOgUIIRCgAToICCEQFhAdEB46BwghEAoQoAFQueIBWKX8AmC4iANoAXABeACAAZMBiAG-GJIBBTExLjE5mAEAoAEBqgEHZ3dzLXdpeg&sclient=psy-ab&ved=0ahUKEwizi6WzzsTpAhVMZ80KHTwBAnUQ4dUDCAw&uact=5

Sampling phenomena are used to save time, cost, and labor, etc. If you have easy access to your respondents (and there is no cost, and time issue) then you should increase the sample size.

You can think about testing a missile, how much sampling you need for this. Similarly there are diseases for which there may be only 1 or two cases in the world. So, first think about your problem and then decide to increase the sample.

Statistics say that there should be at least 2 observations for basic statistics. For ANOVA and modeling related problems you should care about degrees of freedom.

MINIMUM SAMPLE SIZE should be determined before doing research work, not doing the research work and then ask whether the sample is adequate or what sample size should you have.

TEST SAMPLE. The 20 sample you have now may be treated as a test sample. Obtain the descriptive and inferential statistics to calculate the minimum sample size under Cochran's formula:

n = (Z2sigma2) / SE2

where Z = critical value at your specified confidence interval, sigma = *((X^ - µ)/Z)*sqrt(n), and SE = sigma / sqrt(n).

There´s no single answer for sample size, even with an Anova. It´s an inferential procedure and therefore you always must consider the population size, specially if it is a finite one. Significance tests have a lot of problems and criticism (see e.g. Nickerson, 2000; Good & Hardin, 2006).

Hi Danah Alhussain

I think what is most important thing to keep in mind is that whether your sample is representative of the target population. If you don't have idea about the size of population you could take into account factors such as number of variables in your study, number of groups, total number of items on all the measures, sample size of similar studies conducted in past, and then you could apply statistical procedures such as Durbin-Watson, Mahal Distance, normality tests, etc. to ensure that your sample size is adequate and sample is representative as well.

One way ANOVA is just a statistical test which can be used for inferential statistics. The sample size normally depends on the sample size. However, if you use a small sample for the study it is better to use nonparametric tests rather than parametric tests according to your research objectives because if you need to conduct a parametric test, the normality assumption should not be violated. My suggestion is to use a nonparametric test for the analysis according to your objectives.

I agree with @Nandun's answer, non-parametric tests (chi-square) seem better suited for your study.

I agree with dear@José Lucio.

For some researchers, parametric analysiscan be used when size IS more than 30. However, normality test IS réduire After parametric or non-parametric test choisie.