Sampling designs often involve disproportionate sampling across strata. Furthermore, less than 100% of the sample selected typically responds, and some segments of the sample less than others. Consequently, the accuracy of the sample's findings is usually improved by the development and implementation of caseweight adjustments to correct for unequal probabilities of selection, differential response rates across groups, and for possible noncoverage.
Dr. Hembroff has developed the caseweights for hundreds of different phone, web, and mail survey data sets over the past three decades. He has weighted data sets of students or faculty and staff within universities, data sets of residents of cities, counties, states, and the nation. He has also developed caseweights for data sets integrating landline and cell phone samples using the recently developed raking methodology the U.S. Centers for Disease Control and Prevention (CDC) is using for the Behavioral Risk Factor Surveys.
Dr. Hembroff has conducted workshops on the weighting of complex sample data for Michigan State University's faculty and graduate students through MSU's Center for Statistical Consulting (CSTAT) and for other groups.