Abstract:This article presents a novel, general, and effective simulation-inspired approach, called {\it repro samples method}, to conduct statistical inference. The approach studies the performance of artificial samples, referred to as {\it repro samples}, obtained by mimicking the true observed sample to achieve uncertainty quantification and construct confidence sets for parameters of interest with guaranteed coverage rates. Both exact and asymptotic inferences are developed. An attractive feature of the general framework developed is that it does not rely on the large sample central limit theorem and is likelihood-free. As such, it is thus effective for complicated inference problems which we can not solve using the large sample central limit theorem. The proposed method is applicable to a wide range of problems, including many open questions where solutions were previously unavailable, for example, those involving discrete or non-numerical parameters. To reduce the large computational cost of such inference problems, we develop a unique matching scheme to obtain a data-driven candidate set. Moreover, we show the advantages of the proposed framework over the classical Neyman-Pearson framework. We demonstrate the effectiveness of the proposed approach on various models throughout the paper and provide a case study that addresses an open inference question on how to quantify the uncertainty for the unknown number of components in a normal mixture model. To evaluate the empirical performance of our repro samples method, we conduct simulations and study real data examples with comparisons to existing approaches. Although the development pertains to the settings where the large sample central limit theorem does not apply, it also has direct extensions to the cases where the central limit theorem does hold.

Reliable simulation of extremely-truncated log-concave distributions

New Methods of Simulating Lévy Processes

Testing for an ignorable sampling bias under random double truncation

Efficient Sampling from Time-Varying Log-Concave Distributions

Zeroth-Order Sampling Methods for Non-Log-Concave Distributions: Alleviating Metastability by Denoising Diffusion

Maximum Full Likelihood Approach to Randomly Truncated Data

Discrete half-logistic distributions with applications in reliability and risk analysis

Simulating from a gamma distribution with small shape parameter

Nonreversible Langevin Samplers: Splitting Schemes, Analysis and Implementation

Universal Inference Meets Random Projections: A Scalable Test for Log-concavity

Repro Samples Method for Finite- and Large-Sample Inferences

An Exact Method For Simulating Rapidly Decreasing Tempered Stable Distributions

Zero-Truncated Negative Binomial Weighted-Lindley Distribution and Its Application

Efficient Importance Sampling for Rare Event Simulation with Applications

Estimating a Distribution Function with Truncated and Censored Data

Maximum likelihood estimation for left-truncated log-logistic distributions with a given truncation point

Robust Automated Truncation Point Selection for Molecular Simulations

Testing Convex Truncation

Estimating a distribution function for discrete data subject to random truncation with an application to structured finance

The normal law under linear restrictions: simulation and estimation via minimax tilting

Measuring Discrete Risks on Infinite Domains: Theoretical Foundations, Conditional Five Number Summaries, and Data Analyses