Consider a survey where the population being estimated is finite. Using a simple random sample, you draw $n<N$ data. The sampling fraction is $f=\frac{n}{N}$. The variance will be $(1-f)\frac{s^2}{n}$. Can someone explain how the sampling fraction effects this formula?

  • $\begingroup$ Can you clarify your question? Is it about the finite population correction? $\endgroup$ – gung - Reinstate Monica Apr 4 '15 at 1:16
  • $\begingroup$ @gung Yes! How does it affect the variance of a simple random sample? $\endgroup$ – Jnyeboah93 Apr 4 '15 at 1:34
  • $\begingroup$ @gung Do you understand my question? $\endgroup$ – Jnyeboah93 Apr 4 '15 at 1:47
  • $\begingroup$ Did I get it wrong? If so, please edit to correct it. Just make sure it is clear what you are asking. $\endgroup$ – gung - Reinstate Monica Apr 4 '15 at 1:51

In many cases, $f$ is neglected when evaluating the sampling variance. Think about it : if you sample $n = 10 000~$ units among $~N = 300 000 000~$, then $~ f \approx 3 \cdot 10^{-6} \ll 1$.

The most important terms of this formula are : $n$, the sample size, and $s^2$, the variance of the variable you're trying to measure.

Nevertheless, it is somehow logical to see $f$ appear in the formula. Indeed, if $f = 1$ (census), variance of the SRS has to be $0$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.