Haihan Yu
University of Rhode Island, RI·Mathematics & Statistics
Many statistical analyses depend on choosing a model for how the data were generated. In practice that choice is often uncertain, and a wrong model can give unreliable or misleading conclusions.
- Fields
Mathematics & Statistics
- Topics
resampling
bootstrap
subsampling
- Grants
- 1 known grant, through Jul 2028
What they work on
- Develop blended bootstrap techniques that merge subsampling (re-using smaller portions of the data) with the bootstrap (redrawing samples from the data), giving a general strategy for more accurate statistical inference.
- Build empirical likelihood methods for time series and spatial data that combine data transformation and data blocking, two ways of handling dependence between observations, with new bootstrap schemes.
- Establish that these bridged resampling methods are valid and perform well under mild assumptions about the data, without requiring a chosen statistical model.
More about the research
This project aims to build statistical tools that stay valid without strong assumptions about the data. Such tools matter for fields such as environmental science, economics, finance, geology and astronomy, which deal with complex data where model-free methods are especially useful.
Why it matters
Scientists and analysts in environmental science, economics, finance, geology and astronomy often must pick a data model they cannot fully check. Resampling methods that remain valid without a correctly chosen model could make conclusions from time series, spatial and other complex data more reliable. This lowers the risk of misleading results caused by a wrong model choice.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Daniel Nordman
Iowa State University, IA·Mathematics & Statistics
Develop the blended bootstrap, a general strategy that merges the strengths of subsampling and the bootstrap, two different ways of re-using data to measure uncertainty, for more accurate inference on complex data.
Minge Xie
Rutgers University-New Brunswick, NJ·Mathematics & Statistics
Develop PCSI (Predictability-Computability-Stability Inference) for binary and multi-class classification, screening many algorithms by predictive performance and using bootstrap resampling to improve uncertainty estimates, robustness and accuracy.
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.
The school
University of Rhode Island: 9 labs on PI Pages, $137M a year on research, 93 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.