Armeen Taeb
University of Washington, WA·Mathematics & Statistics
When scientists choose a model from data, such as deciding which genes matter or which variables are connected, they want to limit false discoveries: parts of the model that are not really there. For simple choices, a variable is either in or out, so counting false positives is easy.
- Fields
Mathematics & Statistics
Computer Science
- Topics
false discovery rate
false positive control
model selection
- Grants
- 1 known grant, through Jun 2027
What they work on
- 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.
- Determine which properties of the poset structure govern the statistical power and computational complexity of procedures that control false positive error.
- Design model-selection methods with false discovery rate (FDR) guarantees over posets, exploiting specific poset structures to increase power.
- Apply the poset framework to learning rooted phylogenetic trees and to selecting among highly correlated variables with error control.
More about the research
Many modern analyses instead pick structures such as clusters, rankings, causal diagrams, or low-dimensional subspaces. For these, it is unclear what a 'false positive' even means. This project builds a general mathematical way to define and control such errors, so these analyses can be trusted.
Why it matters
Reliable ways to limit false discoveries could make data-driven conclusions more trustworthy when models involve clusters, rankings, causal graphs, or evolutionary trees. The resulting methods could help researchers in computational genomics, image analysis, and the geosciences report findings with stated error guarantees.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Yao Zheng
University of Connecticut, CT·Mathematics & Statistics·Early-career
Develop recurrent neural network (RNN) models for high-dimensional time series that stay interpretable through Granger causality, a test of whether one series predicts another, to balance accuracy and interpretability.
Richard Guo
University of Michigan, MI·Mathematics & Statistics
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.
Adrian Lewis
Cornell University, NY·Mathematics & Statistics
Extend Łojasiewicz-type rescaling techniques to analyze the complexity of first-order and active-set algorithms on semi-algebraic problems in hyperbolic, non-Euclidean and infinite-dimensional spaces.
The school
University of Washington: 81 labs on PI Pages, $1.7B a year on research, 694 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.