Methodology, and the Statistician's Responsibility for BOTH Accuracy AND Relevance

Abstract We should: Measure what is needed for policy guidance, even if it can only be measured poorly. Recognize change as often more important than level. Face the inadequacies of standardization for broad groups, making further corrections. Use more flexible and diversified adjustments, like smear and sweep, dual (smooth and stratumwise) adjustment for one covariate, superstandardization. Scorn statistical maps by learning to use better graphical presentations. Obtain information about residence at fixed intervals before death. Pay attention to specifics such as the proper abscissa for Phillips curves, the economic analogs of physical chemistry, the sampling of student histories in school systems.