McAnulty College and Graduate School of Liberal Arts
Bootstrap, complex surveys, Jackknife, NHANES, variance
The objective of this thesis is to present the various procedures for estimating the variance of specific statistics obtained from different types of survey designs, leading up to more advanced designs such complex surveys. The thesis starts with defining the various sampling designs that are to be used for illustrations (Ch 2). Chapter two gives further descriptions of how various sampling designs are performed (from simple designs to not so simple) and shows the sophistication in calculating the estimates and variances. Chapter three cites the actual equations necessary for estimating the variances of the statistics for each design and demonstrates the potential difficulty especially in estimating the variance of the statistics, as the designs get more complex. Each design is illustrated with numerical examples. Chapter four defines current methods for estimating the variance and introduces the Bootstrap and Jackknife approaches. In Chapter 5 the ideas behind what is considered to be a "complex survey" are described and two nationally known complex surveys (NHANES and NHIS) currently being done in the U.S. are explained as examples. Chapter six reports the main statistical results, comparing the variances, etc., for all the designs and finally a summary conclusion is in chapter 7.
Galiardi, L. (2015). Techniques for Estimating the Variance of Specific Estimators within Complex Surveys (Master's thesis, Duquesne University). Retrieved from https://dsc.duq.edu/etd/563