The National Science Foundation Division of Social and Economic Science awarded William Marsh Rice University $252,672 on October 1, 2025, under the Social, Behavioral, and Economic Sciences program (CFDA 47.075) to develop new test methods for making statistical inference about linear regression coefficients.
The research will produce a complete small-sample and asymptotic theory for randomization-based inference applicable to linear regressions. The methodology is designed to be robust to heteroskedasticity and serial dependence without requiring restrictive assumptions about error-term distributions or sample size. The work frames randomization inference for observational data and Fisher tests for experimental data within a single conceptual framework, establishing connections between the two approaches and suggesting methodological developments. The resulting test statistics will be valid under weaker assumptions than current methods, enabling more precise coefficient estimates and applicability to both observational data analysis and experimental design.
Rice University's Office of Sponsored Projects Division received the award. Performance occurs in Houston, Texas, with a period of performance from October 1, 2025, through July 31, 2027.