An Overview of Bootstrap Resampling
Bootstrap resampling provides a reliable estimate of sampling variation that can be routinely applied in a variety of complex models, such as structural equations, longitudinal observations, and complex sampling schemes. In each case, the bootstrap allows one to investigate the accuracy of models without imposing so many assumptions or pretending that the sampled population has a specific form. All one needs is a clear idea of the sampling mechanism and a fast computer. Starting from some simple examples, this talk maps out the general bootstrapping paradigm, with some pointers to avoid potholes along the way.
Bootstrapping Indices of Residential Segregation
Typically, only point estimates [and not the accompanying standard errors] of indices of urban residential segregation (e.g., Gini and dissimilarity indices) are reported in the professional literature. The reasons for this are varied, ranging from cultural practice to the mathematical intractability of direct solutions. This is unfortunate since substantial variability in these indices often exists across target groups, locations and time. This discussion examines the application of bootstrapping techniques to the analysis of residential segregation across various ethnic groups within Canada’s primary CMAs.
Bootstrap Procedures When Only a Small Number of Simulations is Available.
Conventional procedures for Monte Carlo and bootstrap tests require that B, the number of simulations, satisfy a specific relationship with the level of the test. Otherwise, a test that would instead be exact, will either over-reject or under-reject for finite B. We present expressions for the rejection frequencies associated with existing procedures and propose a new procedure that yields exact Monte Carlo tests for any positive value of B. This procedure, which can also be used for bootstrap tests, is likely to be most useful when simulation is expensive.
Design-based Bootstrapping with Statistics Canada’s Survey Data: The Why’s and How’s
Statistics Canada provides design-based survey bootstrap weights with several of its data files, and encourages their use in the analyses of these data. This talk will discuss the appropriateness of this strategy. It will also address some practical issues in using these weights. The functionality and ease of use of various software tools will also be described.
Why Should You Use the Bootstrap?
Bootstrap procedures are simulation-based techniques which provide estimates of variability, confidence intervals and critical values for tests. The fundamental idea is to create replications by treating the existing data-set as a population from which samples are obtained. In many circumstances bootstrap procedures are simpler to implement than their asymptotic counterparts. In addition, they are often more accurate. This talk will discuss the validity of the bootstrap; why it is often superior to conventional asymptotic methods; and, an empirical application.