[ Wing Tips Home | Contents | Search | Post | Reply | Next | Previous | Up ]
From: Dino Fire, Chapel Hill
Date: 22 Dec 1998
Time: 14:32:34
Of course, the Board should ultimately decide what information they want to collect and analyze. I can be contacted by emailing me at mailto:dinofire@mindspring.com. Again, I will offer my services for free. These include survey design and reporting. Here's the most effective way to do a survey of RWCS members:
1. Use a probability sample. This means that it is not necessary to survey all 7000+ members of the society, as long as each member has an equal and known probability of selection to participate. A sample size of 400 is adequate given the small universe size. This will yield results with a sampling error of +/- 4 percent.
2.) Use a branded mail methodology. This means we select about 700-800 members, at random (actually using n-select methodology), and mail them a survey with a cover letter asking for their participation. The branding part means that RWCS is clearly identified as the sponsor of the survey. Also include the return envelope and postage. Sometimes a dollar bill is included as a premium, but that's probably not necessary here.
3. If 800 people are sent surveys based on a 400 survey design, that implies a "return rate" of about 50%. This is very high, as typical return rates for mail surveys is usually less than 30%. The use of RWCS as the sponsor and the subject matter leads me to believe that at least half will complete the survey. There are other things that help improve return rate...the premium I mentioned before, a follow-up postcard, and live postage are some effective survey treatments.
4. The survey should not be publicized on Wing Tips, the RWCS site, or anywhere else until it's over. This may result in a bias in the data skewed toward members with internet access. It is important that each respondent receive the same survey "treatments."
5. This is a low-cost solution. Printing, postage, and data entry are the only real expenses. A telephone survey is of course an alternative, but far more expensive to execute.