Using survey data, I planed to make log regression. Original sample size is 2000 and there are obaut 10% cases with outcome and 90% withoit it. May I chose amongst thant 90% random sample matced with cases in 1:4 ratio?
I do not know your application, but in general, an adequate random sample size, n, does need to become larger with a larger finite population size, N. However, the fraction n/N becomes smaller with larger N. An important factor in this is standard deviation.
Perhaps you can estimate variance to see when your results seem adequate.