"Power of a study represents the probability of finding a difference that exists in a population".
simply put, it tells how strong is the validity of the finding of your sudy. Power depends on 3 main factor:
significance level, effect size and sample size.
)
significance level is pretty straightforward. We decide it anyway. Thus, it is a no brainer why clinical trials tend to use very low p-levels.
You cant really control effect size. It is an after effect that you only gain after data collection.
But sample size, this is one curious bit. Probability sampling, as its name implies, is a process of taking a sample at random to represent the said population. Once, I heard that at times, probability samplings are better than universal sampling. I wonder, how can a sample of "somethihg" is better than the actual thing? The answer lies in missingness. Missingness is a situation where a data or part of a data is not available. Think of a study about relationship between lifestyle and heart disease. Body Mass Index is important and if it is missing, it will be a problem. when many data is lost from a universal sampling, it reduces the universality of the study, decreasing its power. If the missingness is significant, e.g. more than 30%, it might be better to put them as an exclusion criteria and choose random sampling instead. sure, it will stiffen study generalizability later, but it improves the power.
