Was preparing lecture about experimental inference
Consider an experiment in which the effect of a treatment is measured using a single survey question. Assuming a given effect size, and that we cannot change alpha or the size of the experimental groups, what practical action can we do to increase the power of our experimental design?
Most experiments seem to involve single-item measures of outcomes. This includes field experiments, which are constrained to measuring a single act of voting and survey experiments that measure attitudes using responses to a single question. We know multiple measures are useful (Ansolabehere et al.), but why aren't we using multiple measures in experimental research more often?
Most experiments seem to involve single-item measures of outcomes. This includes field experiments, which are constrained to measuring a single act of voting and survey experiments that measure attitudes using responses to a single question. We know multiple measures are useful (Ansolabehere et al.), but why aren't we using multiple measures in experimental research more often?