Richard Guo
University of Michigan, MI·Mathematics & Statistics
When researchers use the same data both to spot a pattern and to test whether it is real, they inflate the chance of false discoveries. This practice, called "double dipping," undermines how well results replicate in health, economics, political science and other fields.
- Fields
Mathematics & Statistics
Computer Science
- Topics
hypothesis testing
statistical inference
double dipping
- Grants
- 1 known grant, through Aug 2028
What they work on
- Develop a hunt-and-test version of the classical score test for infinite-dimensional parameters such as regression functions and log-densities, using machine learning to find signals and debiasing to keep the test calibrated.
- Remove the randomness of single data splits by aggregating many hunt-and-test replicates and recalibrating them with rank-transformed subsampling, improving power and replicability.
- Speed up derandomized testing and increase its statistical power using low-order approximation techniques and adaptive data aggregation.
More about the research
This project develops statistical tests that can search data flexibly for signals and still check them honestly.
Why it matters
False findings from reusing data weaken trust in research. More powerful and reproducible testing methods could help scientists in health, economics and political science draw conclusions from complex data that hold up on replication, supporting more credible results and greater public trust in research.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Daniela Witten
University of Washington, WA·Mathematics & Statistics
Establish semi-parametric efficiency results showing when, and by how much, extra observations with only features X (no outcome Y) improve inference in semi-supervised learning.
Armeen Taeb
University of Washington, WA·Mathematics & Statistics
Develop a framework that organizes complex model classes (partitions, permutations, directed acyclic graphs, subspaces) as partially ordered sets, and use it to define generalized false positive error.
Matias Cattaneo
Princeton University, NJ·Mathematics & Statistics
Demonstrate the empirical and theoretical limitations of binned regression models applied to mixed-frequency panel data in climate economics, showing when they can lead to incorrect conclusions.
The school
University of Michigan: 124 labs on PI Pages, $2.1B a year on research, 963 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.