0
$\begingroup$

I am sure this question has been answered, and may be obvious to those familiar with online surveys. But I am struggling with it.

I am surveying a specific population (local leaders in a specific state). We are doing an online survey, and will be basing the sample off of a contact list of approximately 3000 emails. I know the response rate is likely to be between 15 and 25%, and I know I want at least 300 complete surveys. This means I need to send the survey to at least 2000 people, which is 2/3 of my sampling frame. Obviously, the entire population (all the people that might be in local leadership roles across the state is somewhat of an "unknown" population)

Is this okay? It seems like a large portion of the sampling frame, and a lot of bias.

The contact list I am using came from an association that is relevant to this sector.

$\endgroup$
2
$\begingroup$

When you have a small population, you can sample all of them; that's not a problem, nor does it introduce a bias.

The bias will come because the people who respond are not a random sample of the population. This is very hard to deal with.

Where the small population becomes an issue is in the analysis. Most statistics are based on an infinite population. When your population is small, you can use the finite population correction:

$$\text{FPC} = \sqrt{\frac{N-n}{N-1}}$$

which, if you get 300 responses becomes $\sqrt{\frac{270}{299}} = 0.95$

$\endgroup$
  • 1
    $\begingroup$ Also worth mentioning: often smaller samples mean more resources available to follow-up non-responders, so increased sample error, but decreased non-sample error. This can be a trade-off worth making. $\endgroup$ – RoryT Apr 6 '18 at 12:09

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.