End users of research often extrapolate from crosstab columns with small sample sizes, despite provided caveats. Unfortunately, such crosstabs overstate differences between subgroups. By adopting Bayesian averages, a technique that e-commerce companies use in product reviews, market researchers can present crosstabs with greater predictive validity. Each column is blended with overall responses to calculate the Bayesian average. A 1,000-record subset of a survey of 6,000 U.S. adults was used to prove the validity of the technique: 60% of the Bayesian averages from the subset were closer to the final results, while only 34% of the traditional crosstabs were closer (6% tied). This simple technique can be applied in Excel and does not require any specialized software, or researchers can use ResearchStory to generate Bayesian crosstabs.
Please join me July 1 for this Insights Associations webinar.