This R Shiny application implements the sequential blocked randomization I wrote with Matthew Kim. In this survey, we replicate Michael Tomz’s 2007 study on audience costs. Respondents first answer 10 demographic questions, are assigned 4 coin-flip confounding treatments, and then assigned to treatment or  ontrol based on those 14 covariates. We successfully replicate Tomz’s results
with less than 20% the sample size.