On the Appropriateness of Bootstrap : Methods to Estimate Standard Errors of Means, Variances, and Median in Complex Sampling Designs

In complex sample design it is difficult to estimate standard of statistic analytically. In some cases there is no neat formula for calculating standard error analytically. Although techniques for variance estimation and confidence intervals exist for complex survey data, they often are cumbersome to implement or do not extend to complex design. Various methods for variance estimation in complex survey data,where the sampling is done without replacement,have been developed so far. These methods include the customary linearization (or taylor) method and resampling method based on Jackknife and balanced repeated replication. These methods all have certain drawbacks, however; the linearization method requires theoritical calculation and subsequent programming of derivatives,which make it cumbersome to implement. It would be desirable to have resampling methods that reuse the existing estimation system repeatedly, using computing power to avoid theoretical work and that can be applied to complex survey design. In this book only bootstrap method for estimating standard error in complex design has been considered.