I'm surveying a population of student in a university, and due to distance and time issues, I cannot go to every campus to survey everybody. This is why I have decided to do an online survey. So I got two options:

  1. I send an open invitation for every (or at least 99%) of the students, thus having the self-selection bias.
  2. I randomly pick everyone from a list and try to reach them through alternative means (like Facebook, e-mail, etc).

For option 1, I'll easily get a larger sample size (I guess around 400); for option 2, I believe I can reach a way shorter sample size (I guess around 100). I'm estimating according to the time I have.

So, in a 95% confidence level I could get a margin of error smaller than 5% in option 1, and around 10% in option 2. But in option 2 I won't have self-selection bias.

What option is the best? And why? Does anyone recommend me anything in a reliable literature source?

  • $\begingroup$ Welcome to stats.SE! Please take a moment to view our tour. You will end up with a sample either way. With that in mind, what attributes are most important to you from your survey respondents? $\endgroup$ – Tavrock Mar 1 '17 at 18:15
  • $\begingroup$ Hey @Tavrock, as I'm surveying them about some preferences, I think the survey respondents must be the most representative of the whole population as possible. I believe non observable attributes might influence the answers, and this is why I actually think the probabilistic one is better (but I'm not sure). Anyway, I need references to the literature to write the foundations of my choice. $\endgroup$ – Bruno Mar 1 '17 at 19:01

Your Answer

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

Browse other questions tagged or ask your own question.