Is there a formula that says, if you ask n questions on a survey, you should give the survey to N people?

I am going to give a 53 item survey to a group of people. I know that there are formulas to determine how many people I should give it to, for probability and confidence; however, is there a formula that would indicate the correct number of people to give it to? This is about the number of items and the number of people. It's related to traditional statistical sample size, but it's about survey items and number of participants. Any help or reference material would be appreciated.

Is there a formula that says, if you ask n questions on a survey, you should give the survey to N people?
I dont see how the number of questions would affect the sample size. The sample size is usually dependant on the total population.





This site can calculate it for you:http://www.surveysystem.com/sscalc.htm
Reply:I work for a statistical agency and I can tell you the only relationship between the number of questions and the sample size is that they both affect the cost of taking the survey. So you may have to make a choice between more questions and more precision (by sampling more), based on your pocketbook.
Reply:No. At least I don't think so.


When you asked 612 people, the answer from the sample is within +- 4% of the true population answer.





From central limit theorem....
Reply:I don't think so. But what you can do is get a feeling for 'internal consistency'. That is, you look at how people who answered one or a certain sample of questions one way, also answered another sample of questions. Did they follow similar patterns or is there no correlation between people and questions?





What this gets at is how consistent they actual survey is. Do people tend to understand what a question is asking? The larger your sample, the better you can demonstrate internal consistency. Another technique related to this is asking questions that are similar in slightly different ways.





Look at how the SAT is analysed to get a feel for this sort of internal consistency. They use it to drop questions that are statistically inconsistent with the rest of the test.





The math for this is to extract certain questions out and redo the overall stats. If you find a question that skews the stats overmuch, then that question is, well, questionable.

web browser