well, that depends on your alpha. so let's break this down to be perfectly correct and not just throw out bullshit for the sake of bull shit. lets talk about trying to determine a proportion for a population, since that seems to be what the topic at hand is. this is the whole formula to determine sample size:
that is a bit cumbersome, so we combine some factors, and estimate the parameter at 1/2 (thats the p-hat and q-hat) and we end up with this:
so now we can solve this. for example (and i haven't looked at the study at hand to see their data, but again, this is just to minimize bullshit spatter) we will use a 95% confidence interval for the mean, which is pretty much the standard, and a width of +/-3%. so Z for an alpha of.025 (half of .05) is 1.96, and W is 2 the width of the proportion.
so that works out to n=1.96^2/.06^2=1067.111
and in order to actually get 1068 respondants, you need to ask more. and if they all respond you have a larger sample size, which means greater accuracy, so with 2500 respondents, there is a very high likelihood that this represents the population.