Confidence intervals for proportions with focus on the US National Health and Nutrition Examination Survey

Xin (Cindy) Feng successfully defended her thesis entitled "Confidence intervals for proportions with focus on the US National Health and Nutrition Examination Survey" on 30 November 2006.
 

It has been noted that the usual confidence interval for proportions does not perform well for large and small values of p, i.e. close to 0 or 1. In surveys, the issue is further complicated by the survey design which can be quite different from a simple random sample. This raises issues of whether to use design effects, effective sample size, or effective degrees of freedom when finding the confidence intervals. There are also many (at least six) different way to compute the confidence interval.

The question is which of the many possible confidence intervals available should be recommended for the US National Health and Nutrition Examination Survey (NHANES) end users and what cautions should be given.

Finally, there are additional complications if the actual interval for the proportion is used in combination with Woodruff's methods to form confidence intervals for small and large quantiles.

This type of interdisciplinary work is a hallmark of our program in Applied Statistics at Simon Fraser University. For more information, please contact Xin Feng (xfeng@sfu.ca) or her supervisor Randy Sitter (sitter@stat.sfu.ca), Department of Statistics and Actuarial Science.

30 November 2006.