Elizabeth Tipton
Professor of Statistics and Data Science
Northwestern University
Faculty Fellow, Institute for Policy Research
I develop methods for designing studies and for building evidence that decision-makers can use.
Recently
- Apr 2026 What schools want vs. what evidence can deliver — PIER seminar, Harvard Graduate School of Education.
- 2025 Elected to the National Academy of Education.
- 2025 Designing Small Evaluation Studies, with Larry Hedges — Sage.
Design Heterogeneity Generalization Synthesis Communication Policy Applications
Seventy papers, 2010–2026. Each block is one paper. Darker blocks are design, heterogeneity, and generalization — the spine of the work. See them all →
The premise
Effects vary. Almost everything hard about evidence follows from that.
If interventions worked the same way for everyone, study design would be simple: one study, one number, true everywhere. Recruitment wouldn’t matter. Synthesis would be averaging.
But effects vary, and everything downstream inherits that. Generalization is hard because effects vary — the units in your study aren’t the only ones a decision will affect and, if they respond differently, you’ve averaged over the wrong set. Evidence fails decision-makers because generalization is hard — the question a superintendent asks is not necessarily the question the trial was designed to answer. And a literature inherits all of this, aggregated as if it had been designed: moderators confounded by history rather than by assignment, so meta-regression estimates something no one intended.
Most of this is decided before anyone measures anything, which is why I work on design: what to build, and for whom. But once you take variation seriously, the analysis has to change too — what an average is estimating, what a variance component means, whether the model can represent a literature with two populations in it rather than one wide one. My lab develops methods on both sides of that line.