stats

Here are a few clarifying questions about the solution: 1) Isn't your approach based on a normal distribution? How did you account for the significant skew in the data? Which distribution did you use/assume? Was the GEV distribution a proper fit and was it used in your calculations? 2) Please define your "no" variable and what is the name of this formula that you used: no =[(zc*s)/E]2? 3) Isn't there a different formula to use for small population sizes to calculate sample sizes? I thought the common formula only applied to large (about 1,000) and didn't change much after the population exceeds around 20,000 (sample sizes stay around 260). 4) Another way to ask question #3 above is, what is a practical solution to having a small population? In our case, if we're collecting student surveys on instructors and we have 50 students and receive surveys from 67% survey response rate, then we cannot have 90% w/ +/-3% nor can we have 90% w/ +/- 5%. We can be within +/-10%, though. So, should we consider ourselves "forced" to adopt a higher MOE for small populations? What do you think about maybe making these 2 rules: i) we do not accept any sample sizes below 30 (which I recall is a rule of thumb anyway for normal distributions). ii) we operate with 10% MOE until we reach at least 150 population, at which point, we can narrow our MOE to 5%?