Multicalibration for Unbiased Model-Based Prevalence Estimation
Two empirical applications -- estimating employment prevalence across U.S. states using the American Community Survey and classifying political texts across four countries using an LLM -- demonstrate that multicalibration substantially reduces bias in practice, while highlighting that calibration data should cover the...
Fridolin Linder, Thomas J. Leeper, Daniel Haimovich et al.
· 1 citation