Year: 2026
Type: preprint
Abstract: Researchers often observe group quantiles but not group means, raising the question of whether these summaries alone support conventional regression analysis. We develop estimation and inference methods for linear models of group means under quantile-only and mixed disclosure. We establish conditions on disclosure design for consistent estimation and asymptotic normality at the parametric rate. Under fixed disclosure, we characterize sharp identified sets and construct simultaneous confidence regions, clarifying how informativeness depends on the number and placement of disclosed quantiles. Applications to education, social care, and medical expenditure illustrate the methods. The R package qonlyreg implements the proposed procedures.
Source: arXiv (Cornell University)
Authors: Iwasawa, Masamune, Nakanishi, Hayato
Institutions: Doshisha University, Kanagawa University
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7223107441