arXiv subjects

Xinran Miao

Publications and source records attributed to Xinran Miao.

5 recordsLinked to original sources

arXiv2025

SLOPE and Designing Robust Studies for Generalization

A popular task in generalization is to learn about a new, target population based on data from an existing, source population. This task relies on conditional exchangeability, which asserts that differences between the source and target populations are fully captured by observable characteristics of the two populations. Unfortunately, this assumption is often untenable in practice due to unobservable differences between the source and target populations. Worse, the assumption cannot be verified with data, warranting the need for robust data collection processes and study designs that are inherently less sensitive to violation of the assumption. In this paper, we propose SLOPE (Sensitivity of LOcal Perturbations from Exchangeability), a simple, intuitive, and novel measure that quantifies the sensitivity to local violation of conditional exchangeability. SLOPE combines ideas from sensitivity analysis in causal inference and derivative-based measure of robustness from Hampel (1974). Among other properties, SLOPE can help investigators to choose (a) a robust source or target population or (b) a robust estimand. Also, we show an analytic relationship between SLOPE and influence functions, which investigators can use to derive SLOPE given an influence function. We conclude with a re-analysis of a multi-national randomized experiment and illustrate the role of SLOPE in informing robust study designs for generalization.

stat.ME↗

arXiv2024

Transporting causal effects from a randomized trial without "transportability:" a case study of political advertising during U.S. elections

During the 2020 U.S. presidential election, Aggarwal et al. (2023) conducted a large-scale randomized experiment to evaluate a digital ad campaign against Trump in five battleground states. While the study found no effect on voter turnout, it's unclear whether this null result generalizes to other battleground states, notably Georgia, which played a unique role in the 2020 election and differs from the battleground states. Inspired by the study, we present a transfer learning framework to estimate treatment effects in a target population (e.g., Georgia) based on a randomized experiment from a source population (e.g., the five battleground states). Our framework is based on a sensitivity analysis that allows for violation of transportability, a popular yet impractical assumption which requires all differences between the source and target populations to be characterized by observed variables. Under our framework, we propose two estimators of the target treatment effect: a simple regression estimator with bootstrap, which we recommend for practitioners in this field, and an estimator based on the efficient influence function. Importantly, both estimators allow for covariates to differ between the target and the source populations, another common scenario in practice. We also propose a new, sample splitting approach to calibrate the sensitivity parameter. We apply our framework to estimate the effect of the ad campaign on voter turnout in Georgia during the 2020 election. Our findings indicate that small departures from transportability can lead to dramatically different ad effects across counties of Georgia. The direction of the effects is largely driven by racial composition: counties with higher White and lower Black percents tend to show positive effects, while counties with higher Latinx percents tend to show negative effects.

stat.AP↗

arXiv2024

Assumption-Lean and Data-Adaptive Post-Prediction Inference

A primary challenge facing modern scientific research is the limited availability of gold-standard data which can be costly, labor-intensive, or invasive to obtain. With the rapid development of machine learning (ML), scientists can now employ ML algorithms to predict gold-standard outcomes with variables that are easier to obtain. However, these predicted outcomes are often used directly in subsequent statistical analyses, ignoring imprecision and heterogeneity introduced by the prediction procedure. This will likely result in false positive findings and invalid scientific conclusions. In this work, we introduce PoSt-Prediction Adaptive inference (PSPA) that allows valid and powerful inference based on ML-predicted data. Its "assumption-lean" property guarantees reliable statistical inference without assumptions on the ML prediction. Its "data-adaptive" feature guarantees an efficiency gain over existing methods, regardless of the accuracy of ML prediction. We demonstrate the statistical superiority and broad applicability of our method through simulations and real-data applications.

stat.ME↗

arXiv2022

Interactive Visualization and Representation Analysis Applied to Glacier Segmentation

Interpretability has attracted increasing attention in earth observation problems. We apply interactive visualization and representation analysis to guide interpretation of glacier segmentation models. We visualize the activations from a U-Net to understand and evaluate the model performance. We build an online interface using the Shiny R package to provide comprehensive error analysis of the predictions. Users can interact with the panels and discover model failure modes. Further, we discuss how visualization can provide sanity checks during data preprocessing and model training.

cs.CV↗

arXiv2022

Source data selection for out-of-domain generalization

Models that perform out-of-domain generalization borrow knowledge from heterogeneous source data and apply it to a related but distinct target task. Transfer learning has proven effective for accomplishing this generalization in many applications. However, poor selection of a source dataset can lead to poor performance on the target, a phenomenon called negative transfer. In order to take full advantage of available source data, this work studies source data selection with respect to a target task. We propose two source selection methods that are based on the multi-bandit theory and random search, respectively. We conduct a thorough empirical evaluation on both simulated and real data. Our proposals can be also viewed as diagnostics for the existence of a reweighted source subsamples that perform better than the random selection of available samples.

cs.LG↗