WesVar

WesVar is a Windows-based software package for analyzing data collected by surveys with complex sample designs. Sample designs that can be analyzed include stratified or unstratified designs, single- or multi-stage designs, one-time or longitudinal designs. A key aspect of WesVar is the [Page 961]calculation of design-based standard errors that take the complex sample design (e.g., stratification and clustering) into account.

WesVar calculates sampling variances using the jackknife and balanced repeated replication methods. WesVar users can calculate and use their own replicate weights or have WesVar calculate the replicate weights. WesVar can perform weight adjustments for nonresponse, post-stratification, and raking. Such weight adjustments are performed on both the full-sample weights and the replicate weights, so that calculated sampling variances take into account the effects of weight adjustments.

The estimates and standard errors calculated by WesVar take into account the sample design and adjustments to sampling weights. Statistical software that ignores the sample design and assumes simple random sampling typically underestimates standard errors.

WesVar can calculate all of the following:

• Multi-way tables containing totals, means, or proportions, along with estimated standard error, coefficient of variations, confidence intervals, and design effects
• Estimates of medians and other quantiles, along with estimated standard errors
• Complex functions of estimates, such as ratios, differences of ratios, and log-odds ratios, along with estimated standard errors
• Chi-square tests of independence for two-way tables of weighted frequencies
• Estimated coefficients and their standard errors for linear and logistic regression models and associated significance tests for linear combinations of parameters

WesVar was developed and is distributed by Westat. The program and user manual can be downloaded from the WesVar Web site.