[Link to discussion points][1] In an era of increasing focus on preregistration and methodological rigor, design analysis—the analysis of how well a proposed design can answer the research questions—is invaluable for ensuring that we can actually draw meaningful inferences from education research. Indeed, major funding agencies such as the Institute of Education Sciences require at least a power analysis. However, these analyses are generally perfunctory and superficial, especially for complex designs (e.g. mediation). These superficial techniques rely on rules of thumb and unrealistic assumptions, and only address statistical power without addressing other important issues such as assumption violations, confidence interval coverage, validity of Bayesian inference, bias in estimating causal parameters, etc. Preregistering a poor and underpowered design does not improve or diagnose the design. More thoughtful design analysis is possible, for instance using Monte Carlo techniques, but there are a variety of barriers to adoption. In this unconference session, we’ll discuss these barriers and how we can improve the quality of study design and design analysis. Among the barriers we’ll discuss are 1) the knowledge and skills that the researcher brings to the study, including well-documented researcher misconceptions about statistical inference (e.g. Kahneman & Tversky, 1971), 2) lack of resources (time, money, and expertise) for performing an adequate design analysis, 3) difficulty of currently available software solutions for performing design analysis, and 4) incentive structures of grants and institutions that do not fully reward rigorous design analysis. As a group, we can brainstorm and try to move towards a plan of action to support this important work before the study begins. [1]: https://pad.riseup.net/p/osperpower