Geraldine - very difficult to state. It depends on certain factors. Do you want to/need to generalise? If that is the case, you need to have an approximate idea of the total population and then you could perform a power calculation to predict the minimum sample size. In absence of that, then the required number is subjective and most likely to be 'as many as you can get'. You might state a subjective minimum i.e. 50 that you feel you could do anything 'meaningful' in terms of interpretation of the collected data.

Whatever the number, be aware that recruiting hard to access populations is usually a long and laborious task - and may have to employ techniques such as snowball sampling.

Thank you Dean. Actually I do not need to generalise. I will be doing a cross-cultural, cross-national comparative study and since my data collection will take place in two developing countries with two different refugee populations I am quite puzzled about what would constitute a good sample. Naturally the more respondents I can get the better, but that subjective minimum that you are talking about, let's say I get 100 respondents in each locations, would it still have good scientific value ? Obviously, the value of the research pertains to many factors, but if isolating that one, the sample size, would it still be scientifically acceptable ? Thank you !

Geraldine - I think you want an acceptable sample size when you are trying to survey a population that is hard to access, especially if you don't have access to data that indicate the size of that population in the first place. It means you are dealing with infinite population. The number of units in a finite population is denoted by N which is the size of the population. In infinite population, sometimes it is not possible to count the units contained in the population. In order to calculate the minimum sample size, you can consider the standard normal deviation set at 95% confidence level (1.96), percentage picking a choice or response (50% = 0.5) and the confidence interval (0.05 = ±5). The formula is: n = z 2  (p)(1-p)              c 2 

Where: z = standard normal deviation set at 95% confidence level p = percentage picking a choice or response c = confidence interval interval for your estimation 

The formula for estimating sample size is not that strongly affected by the population N in a large population, such as when you are dealing with a population N of over a thousand.

You can find several sample size calculators on the internet, which follow the standard formula described by Manzoor Hussain. Try them out by putting different large values into the population size portion of the calculator, and you will see what I mean