These guidelines provide direction for analysis of Survey of the Health of Wisconsin (SHOW) data. They explain the complex sample survey and weighting methodology, and provide advice on how to use sample weights for descriptive statistics, as well as for more complex analyses.
SHOW uses a complex, probability sampling approach (see sampling strategy in survey overview) when identifying households to visit each year. This is in order to meet SHOW’s goals in providing representative data for following health and health determinants in Wisconsin.
In order to provide unbiased estimates for the target population, SHOW generates sample weights that accompany each data point. These reflect the sample design (probability of selection), adjustment for nonresponse and post-stratification adjustment to bring the sample back to being representative of the target population based on a few demographic variables. Sample weights were generated for each person for each phase of data collection: the in-home questionnaire visit, the self-administered questionnaire phase and the sample collection visit.
In order to generalize findings to the target population and for proper, unbiased estimates and standard errors of estimates of the target population, sampling weights and the stratification and clustering of SHOW’s survey design should be incorporated into the analysis. Ignoring the complex design can lead to biased estimates and overstated significance levels. More details on guidance are available from the SHOW Data Team, including syntax examples for incorporating weights and design features in certain procedures in SAS, R, and Stata.
SHOW in 2017 is following individuals over-time, therefore longitudinal analysis for a subset of SHOW adults with prior participation in 2008-13 will also be available.
There may be some analyses where sampling weights are not necessary. For additional questions about using SHOW data, please contact the SHOW data team at firstname.lastname@example.org.
For additional questions or for guidance on using SHOW data, you can view our FAQs document. If you still have questions about accessing and analyzing SHOW data, please contact the SHOW data team at email@example.com.