arXiv subjects

Reagan Mozer

Publications and source records attributed to Reagan Mozer.

6 recordsLinked to original sources

arXiv2026

PPI is the Difference Estimator: Recognizing the Survey Sampling Roots of Prediction-Powered Inference

Prediction-powered inference (PPI) is a rapidly growing framework for combining machine learning predictions with a small set of gold-standard labels to conduct valid statistical inference. In this article, I argue that the core estimators underlying PPI are equivalent to well-established estimators from the survey sampling literature dating back to the 1970s. Specifically, the PPI estimator for a population mean is algebraically equivalent to the difference estimator of Cassel et al. (1976), and PPI plus corresponds to the generalized regression (GREG) estimator of Sarndal et al. (2003). Recognizing this equivalence, I consider what part of PPI is inherited from a long-standing literature in statistics, what part is genuinely new, and where inferential claims require care. After introducing the two frameworks and establishing their equivalence, I break down where PPI diverges from model-assisted estimation, including differences in the mode of inference, the role of the unlabeled data pool, and the consequences of differential prediction error for subgroup estimands such as the average treatment effect. I then identify what each framework offers the other: PPI researchers can draw on the survey sampling literature's well-developed theory of calibration, optimal allocation, and design-based diagnostics, while survey sampling researchers can benefit from PPI's extensions to non-standard estimands and its accessible software ecosystem. The article closes with a call for integration between these two communities, motivated by the growing use of large language models as measurement instruments in applied research.

stat.ME↗

arXiv2026

Stratified Sampling for Model-Assisted Estimation with Surrogate Outcomes

In many randomized trials, outcomes such as essays or open-ended responses must be manually scored as a preliminary step to impact analysis, a process that is costly and limiting. Model-assisted estimation offers a way to combine surrogate outcomes generated by machine learning or large language models with a human-coded subset, yet typical implementations use simple random sampling and therefore overlook systematic variation in surrogate prediction error. We extend this framework by incorporating stratified sampling to more efficiently allocate human coding effort. We derive the exact variance of the stratified model-assisted estimator, characterize conditions under which stratification improves precision, and identify a Neyman-type optimal allocation rule that oversamples strata with larger residual variance. We evaluate our methods through a comprehensive simulation study to assess finite-sample performance. Overall, we find stratification consistently improves efficiency when surrogate prediction errors exhibit structured bias or heteroskedasticity. We also present two empirical applications, one using data from an education RCT and one using a large observational corpus, to illustrate how these methods can be implemented in practice using ChatGPT-generated surrogate outcomes. Overall, this framework provides a practical design-based approach for leveraging surrogate outcomes and strategically allocating human coding effort to obtain unbiased estimates with greater efficiency. While motivated by text-as-data applications, the methodology applies broadly to any setting where outcome measurement is costly or prohibitive, and can be applied to comparisons across groups or estimating the mean of a single group.

stat.ME↗

arXiv2024

Leveraging text data for causal inference using electronic health records

In studies that rely on data from electronic health records (EHRs), unstructured text data such as clinical progress notes offer a rich source of information about patient characteristics and care that may be missing from structured data. Despite the prevalence of text in clinical research, these data are often ignored for the purposes of quantitative analysis due their complexity. This paper presents a unified framework for leveraging text data to support causal inference with electronic health data at multiple stages of analysis. In particular, we consider how natural language processing and statistical text analysis can be combined with standard inferential techniques to address common challenges due to missing data, confounding bias, and treatment effect heterogeneity. Through an application to a recent EHR study investigating the effects of a non-randomized medical intervention on patient outcomes, we show how incorporating text data in a traditional matching analysis can help strengthen the validity of an estimated treatment effect and identify patient subgroups that may benefit most from treatment. We believe these methods have the potential to expand the scope of secondary analysis of clinical data to domains where structured EHR data is limited, such as in developing countries. To this end, we provide code and open-source replication materials to encourage adoption and broader exploration of these techniques in clinical research.

cs.CL↗

arXiv2023

More power to you: Using machine learning to augment human coding for more efficient inference in text-based randomized trials

For randomized trials that use text as an outcome, traditional approaches for assessing treatment impact require that each document first be manually coded for constructs of interest by trained human raters. This process, the current standard, is both time-consuming and limiting: even the largest human coding efforts are typically constrained to measure only a small set of dimensions across a subsample of available texts. In this work, we present an inferential framework that can be used to increase the power of an impact assessment, given a fixed human-coding budget, by taking advantage of any "untapped" observations -- those documents not manually scored due to time or resource constraints -- as a supplementary resource. Our approach, a methodological combination of causal inference, survey sampling methods, and machine learning, has four steps: (1) select and code a sample of documents; (2) build a machine learning model to predict the human-coded outcomes from a set of automatically extracted text features; (3) generate machine-predicted scores for all documents and use these scores to estimate treatment impacts; and (4) adjust the final impact estimates using the residual differences between human-coded and machine-predicted outcomes. This final step ensures any biases in the modeling procedure do not propagate to biases in final estimated effects. Through an extensive simulation study and an application to a recent field trial in education, we show that our proposed approach can be used to reduce the scope of a human-coding effort while maintaining nominal power to detect a significant treatment impact.

stat.ME↗

arXiv2019

Bayesian analysis of longitudinal studies with treatment by indication

It is often of interest in observational studies to measure the causal effect of a treatment on time-to-event outcomes. In a medical setting, observational studies commonly involve patients who initiate medication therapy and others who do not, and the goal is to infer the effect of medication therapy on time until recovery, a pre-defined level of improvement, or some other time-to-event outcome. A difficulty with such studies is that the notion of a medication initiation time does not exist in the control group. We propose an approach to infer causal effects of an intervention in longitudinal observational studies when the time of treatment assignment is only observed for treated units and where treatment is given by indication. We present a framework for conceptualizing an underlying randomized experiment in this setting based on separating the process that governs the time of study arm assignment from the mechanism that determines the assignment. Our approach involves inferring the missing times of assignment followed by estimating treatment effects. This approach allows us to incorporate uncertainty about the missing times of study arm assignment, which induces uncertainty in both the selection of the control group and the measurement of time-to-event outcomes for these controls. We demonstrate our approach to study the effects on mortality of inappropriately prescribing phosphodiesterase type 5 inhibitors (PDE5Is), a medication contraindicated for groups 2 and 3 pulmonary hypertension, using administrative data from the Veterans Affairs (VA) health care system.

stat.ME↗

arXiv2019

Matching with Text Data: An Experimental Evaluation of Methods for Matching Documents and of Measuring Match Quality

Matching for causal inference is a well-studied problem, but standard methods fail when the units to match are text documents: the high-dimensional and rich nature of the data renders exact matching infeasible, causes propensity scores to produce incomparable matches, and makes assessing match quality difficult. In this paper, we characterize a framework for matching text documents that decomposes existing methods into: (1) the choice of text representation, and (2) the choice of distance metric. We investigate how different choices within this framework affect both the quantity and quality of matches identified through a systematic multifactor evaluation experiment using human subjects. Altogether we evaluate over 100 unique text matching methods along with 5 comparison methods taken from the literature. Our experimental results identify methods that generate matches with higher subjective match quality than current state-of-the-art techniques. We enhance the precision of these results by developing a predictive model to estimate the match quality of pairs of text documents as a function of our various distance scores. This model, which we find successfully mimics human judgment, also allows for approximate and unsupervised evaluation of new procedures. We then employ the identified best method to illustrate the utility of text matching in two applications. First, we engage with a substantive debate in the study of media bias by using text matching to control for topic selection when comparing news articles from thirteen news sources. We then show how conditioning on text data leads to more precise causal inferences in an observational study examining the effects of a medical intervention.

stat.ME↗