The Complex Sample Bag of Little Bootstraps (303216)

Keywords: complex sample,csBLB,bootstrap,BLB,complex sampling,complex survey,survey,bag of little bootstraps

Complex sampling designs are becoming more prevalent, especially in survey research studies. The designs let the researcher analyze samples across sampling stages. They ensure sufficient representation of sub-population groups, while being cost-effective and precise. However, with complex sampling comes complex statistical analysis, and that analysis is rendered more computationally inefficient due to the increasing prevalence of large samples. The complex sample classical bootstrap provides better quality assessment estimators than the asymptotic approximation, but this method requires significant processing time and computer memory with large complex data sets. The Bag of Little Bootstraps (BLB) is a method invented to produce the same statistical properties as the classical bootstrap, but more computationally efficient with large samples. It achieves this by resampling from smaller subsamples of the data set. The poster will introduce the complex sample Bag of Little Bootstraps (csBLB), a BLB mutation that shares the same properties and efficiency while tailored for complex samples. The algorithm, hyperparameters, simulations, and theoretical analysis will be given in the poster.