arXiv subjects

Walter Dempsey

Publications and source records attributed to Walter Dempsey.

At least 37 records · Page 2Linked to original sources

arXiv2023

Design of Experiments with Sequential Randomizations on Multiple Timescales: The Hybrid Experimental Design

Psychological interventions, especially those leveraging mobile and wireless technologies, often include multiple components that are delivered and adapted on multiple timescales (e.g., coaching sessions adapted monthly based on clinical progress, combined with motivational messages from a mobile device adapted daily based on the person's daily emotional state). The hybrid experimental design (HED) is a new experimental approach that enables researchers to answer scientific questions about the construction of psychological interventions in which components are delivered and adapted on different timescales. These designs involve sequential randomizations of study participants to intervention components, each at an appropriate timescale (e.g., monthly randomization to different intensities of coaching sessions and daily randomization to different forms of motivational messages). The goal of the current manuscript is twofold. The first is to highlight the flexibility of the HED by conceptualizing this experimental approach as a special form of a factorial design in which different factors are introduced at multiple timescales. We also discuss how the structure of the HED can vary depending on the scientific question(s) motivating the study. The second goal is to explain how data from various types of HEDs can be analyzed to answer a variety of scientific questions about the development of multi-component psychological interventions. For illustration we use a completed HED to inform the development of a technology-based weight loss intervention that integrates components that are delivered and adapted on multiple timescales.

stat.ME↗

arXiv2022

Estimating Time-Varying Direct and Indirect Causal Excursion Effects with Longitudinal Binary Outcomes

Construction of just-in-time adaptive interventions, such as prompts delivered by mobile apps to promote and maintain behavioral change, requires knowledge about time-varying moderated effects to inform when and how we deliver intervention options. Micro-randomized trials (MRT) have emerged as a sequentially randomized design to gather requisite data for effect estimation. The existing literature (Qian et al., 2020; Boruvka et al., 2018; Dempsey et al., 2020) has defined a general class of causal estimands, referred to as "causal excursion effects", to assess the time-varying moderated effect. However, there is limited statistical literature on how to address potential between-cluster treatment effect heterogeneity and within-cluster interference in a sequential treatment setting for longitudinal binary outcomes. In this paper, based on a cluster conceptualization of the potential outcomes, we define a larger class of direct and indirect causal excursion effects for proximal and lagged binary outcomes, and propose a new inferential procedure that addresses effect heterogeneity and interference. We provide theoretical guarantees of consistency and asymptotic normality of the estimator. Extensive simulation studies confirm our theory empirically and show the proposed procedure provides consistent point estimator and interval estimates with valid coverage. Finally, we analyze a data set from a multi-institution MRT study to assess the time-varying moderated effects of mobile prompts upon binary study engagement outcomes.

stat.ME↗

arXiv2022

Addressing selection bias and measurement error in COVID-19 case count data using auxiliary information

Coronavirus case-count data has influenced government policies and drives most epidemiological forecasts. Limited testing is cited as the key driver behind minimal information on the COVID-19 pandemic. While expanded testing is laudable, measurement error and selection bias are the two greatest problems limiting our understanding of the COVID-19 pandemic; neither can be fully addressed by increased testing capacity. In this paper, we demonstrate their impact on estimation of point prevalence and the effective reproduction number. We show that estimates based on the millions of molecular tests in the US has the same mean square error as a small simple random sample. To address this, a procedure is presented that combines case-count data and random samples over time to estimate selection propensities based on key covariate information. We then combine these selection propensities with epidemiological forecast models to construct a \emph{doubly robust} estimation method that accounts for both measurement-error and selection bias. This method is then applied to estimate Indiana's active infection prevalence using case-count, hospitalization, and death data with demographic information, a statewide random molecular sample collected from April 25--29th, and Delphi's COVID-19 Trends and Impact Survey. We end with a series of recommendations based on the proposed methodology.

stat.ME↗

arXiv2022

Node-level community detection within edge exchangeable models for interaction processes

Scientists are increasingly interested in discovering community structure from modern relational data arising on large-scale social networks. While many methods have been proposed for learning community structure, few account for the fact that these modern networks arise from processes of interactions in the population. We introduce block edge exchangeable models (BEEM) for the study of interaction networks with latent node-level community structure. The block vertex components model (B-VCM) is derived as a canonical example. Several theoretical and practical advantages over traditional vertex-centric approaches are highlighted. In particular, BEEMs allow for sparse degree structure and power-law degree distributions within communities. Our theoretical analysis bounds the misspecification rate of block assignments, while supporting simulations show the properties of the network can be recovered. A computationally tractable Gibbs algorithm is derived. We demonstrate the proposed model using post-comment interaction data from Talklife, a large-scale online peer-to-peer support network, and contrast the learned communities from those using standard algorithms including spectral clustering and degree-correct stochastic block models.

stat.ME↗

arXiv2022

Recurrent event analysis in the presence of real-time high frequency data via random subsampling

Digital monitoring studies collect real-time high frequency data via mobile sensors in the subjects' natural environment. This data can be used to model the impact of changes in physiology on recurrent event outcomes such as smoking, drug use, alcohol use, or self-identified moments of suicide ideation. Likelihood calculations for the recurrent event analysis, however, become computationally prohibitive in this setting. Motivated by this, a random subsampling framework is proposed for computationally efficient, approximate likelihood-based estimation. A subsampling-unbiased estimator for the derivative of the cumulative hazard enters into an approximation of log-likelihood. The estimator has two sources of variation: the first due to the recurrent event model and the second due to subsampling. The latter can be reduced by increasing the sampling rate; however, this leads to increased computational costs. The approximate score equations are equivalent to logistic regression score equations, allowing for standard, "off-the-shelf" software to be used in fitting these models. Simulations demonstrate the method and efficiency-computation trade-off. We end by illustrating our approach using data from a digital monitoring study of suicidal ideation.

stat.ME↗

arXiv2021

Assessing Time-Varying Causal Effect Moderation in the Presence of Cluster-Level Treatment Effect Heterogeneity

The micro-randomized trial (MRT) is a sequential randomized experimental design to empirically evaluate the effectiveness of mobile health (mHealth) intervention components that may be delivered at hundreds or thousands of decision points. MRTs have motivated a new class of causal estimands, termed "causal excursion effects", for which semiparametric inference can be conducted via a weighted, centered least squares criterion (Boruvka et al., 2018). Existing methods assume between-subject independence and non-interference. Deviations from these assumptions often occur. In this paper, causal excursion effects are revisited under potential cluster-level treatment effect heterogeneity and interference, where the treatment effect of interest may depend on cluster-level moderators. Utility of the proposed methods is shown by analyzing data from a multi-institution cohort of first year medical residents in the United States.

stat.ME↗

arXiv2021

Kernel Deformed Exponential Families for Sparse Continuous Attention

Attention mechanisms take an expectation of a data representation with respect to probability weights. This creates summary statistics that focus on important features. Recently, (Martins et al. 2020, 2021) proposed continuous attention mechanisms, focusing on unimodal attention densities from the exponential and deformed exponential families: the latter has sparse support. (Farinhas et al. 2021) extended this to use Gaussian mixture attention densities, which are a flexible class with dense support. In this paper, we extend this to two general flexible classes: kernel exponential families and our new sparse counterpart kernel deformed exponential families. Theoretically, we show new existence results for both kernel exponential and deformed exponential families, and that the deformed case has similar approximation capabilities to kernel exponential families. Experiments show that kernel deformed exponential families can attend to multiple compact regions of the data domain.

cs.LG↗

arXiv2020

A Robust Functional EM Algorithm for Incomplete Panel Count Data

Panel count data describes aggregated counts of recurrent events observed at discrete time points. To understand dynamics of health behaviors, the field of quantitative behavioral research has evolved to increasingly rely upon panel count data collected via multiple self reports, for example, about frequencies of smoking using in-the-moment surveys on mobile devices. However, missing reports are common and present a major barrier to downstream statistical learning. As a first step, under a missing completely at random assumption (MCAR), we propose a simple yet widely applicable functional EM algorithm to estimate the counting process mean function, which is of central interest to behavioral scientists. The proposed approach wraps several popular panel count inference methods, seamlessly deals with incomplete counts and is robust to misspecification of the Poisson process assumption. Theoretical analysis of the proposed algorithm provides finite-sample guarantees by expanding parametric EM theory to our general non-parametric setting. We illustrate the utility of the proposed algorithm through numerical experiments and an analysis of smoking cessation data. We also discuss useful extensions to address deviations from the MCAR assumption and covariate effects.

stat.ME↗

arXiv2019

Hierarchical network models for structured exchangeable interaction processes

Network data often arises via a series of structured interactions among a population of constituent elements. E-mail exchanges, for example, have a single sender followed by potentially multiple receivers. Scientific articles, on the other hand, may have multiple subject areas and multiple authors. We introduce hierarchical edge exchangeable models for the study of these structured interaction networks. In particular, we introduce the hierarchical vertex components model as a canonical example, which partially pools information via a latent, shared population-level distribution. Theoretical analysis and supporting simulations provide clear model interpretation, and establish global sparsity and power-law degree distribution. A computationally tractable Gibbs algorithm is derived. We demonstrate the model on both the Enron e-mail dataset and an ArXiv dataset, showing goodness of fit of the model via posterior predictive validation.

stat.ME↗

arXiv2018

Exchangeable, Markov multi-state survival process

We consider exchangeable Markov multi-state survival processes -- temporal processes taking values over a state-space$\mathcal{S}$ with at least one absorbing failure state $\flat \in \mathcal{S}$ that satisfy natural invariance properties of exchangeability and consistency under subsampling. The set of processes contains many well-known examples from health and epidemiology -- survival, illness-death, competing risk, and comorbidity processes; an extension leads to recurrent event processes. We characterize exchangeable Markov multi-state survival processes in both discrete and continuous time. Statistical considerations impose natural constraints on the space of models appropriate for applied work. In particular, we describe constraints arising from the notion of composable systems. We end with an application of the developed models to irregularly sampled and potentially censored multi-state survival data, developing a Markov chain Monte Carlo algorithm for posterior computation.

stat.ME↗

arXiv2017

The stratified micro-randomized trial design: sample size considerations for testing nested causal effects of time-varying treatments

Technological advancements in the field of mobile devices and wearable sensors have helped overcome obstacles in the delivery of care, making it possible to deliver behavioral treatments anytime and anywhere. Increasingly the delivery of these treatments is triggered by predictions of risk or engagement which may have been impacted by prior treatments. Furthermore the treatments are often designed to have an impact on individuals over a span of time during which subsequent treatments may be provided. Here we discuss our work on the design of a mobile health smoking cessation experimental study in which two challenges arose. First the randomizations to treatment should occur at times of stress and second the outcome of interest accrues over a period that may include subsequent treatment. To address these challenges we develop the "stratified micro-randomized trial," in which each individual is randomized among treatments at times determined by predictions constructed from outcomes to prior treatment and with randomization probabilities depending on these outcomes. We define both conditional and marginal proximal treatment effects. Depending on the scientific goal these effects may be defined over a period of time during which subsequent treatments may be provided. We develop a primary analysis method and associated sample size formulae for testing these effects.

stat.AP↗

arXiv2017

Relational exchangeability

A relationally exchangeable structure is a random combinatorial structure whose law is invariant with respect to relabeling its relations, as opposed to its elements. Aside from exchangeable random partitions, examples include edge exchangeable random graphs and hypergraphs, path exchangeable processes, and a range of other network-like structures that arise in statistical applications. We prove a de Finetti-type structure theorem for the general class of relationally exchangeable structures.

math.ST↗

arXiv2016

A framework for statistical network modeling

Basic principles of statistical inference are commonly violated in network data analysis. Under the current approach, it is often impossible to identify a model that accommodates known empirical behaviors, possesses crucial inferential properties, and accurately models the data generating process. In the absence of one or more of these properties, sensible inference from network data cannot be assured. Our proposed framework decomposes every network model into a (relatively) exchangeable data generating process} and a sampling mechanism that relates observed data to the population network. This framework, which encompasses all models in current use as well as many new models, such as edge exchangeable and relationally exchangeable models, that lie outside the existing paradigm, offers a sound context within which to develop theory and methods for network analysis.

math.ST↗

arXiv2016

Edge exchangeable models for network data

Exchangeable models for countable vertex-labeled graphs cannot replicate the large sample behaviors of sparsity and power law degree distribution observed in many network datasets. Out of this mathematical impossibility emerges the question of how network data can be modeled in a way that reflects known empirical behaviors and respects basic statistical principles. We address this question by observing that edges, not vertices, act as the statistical units in networks constructed from interaction data, making a theory of edge-labeled networks more natural for many applications. In this context we introduce the concept of {\em edge exchangeability}, which unlike its vertex exchangeable counterpart admits models for networks with sparse and/or power law structure. Our characterization of edge exchangeable networks gives rise to a class of nonparametric models, akin to graphon models in the vertex exchangeable setting. Within this class, we identify a tractable family of distributions with a clear interpretation and suitable theoretical properties, whose significance in estimation, prediction, and testing we demonstrate.

math.ST↗

arXiv2016

Vital variables and survival processes

The focus of a survival study is partly on the distribution of survival times, and partly on the health or quality of life of patients while they live. Health varies over time, and survival is the most basic aspect of health, so the two aspects are closely intertwined. Depending on the nature of the study, a range of variables may be measured; some constant in time, others not; some regarded as responses, others as explanatory risk factors; some directly and personally health-related, others less directly so. This paper begins by classifying variables that may arise in such a setting, emphasizing in particular, the mathematical distinction between vital and non-vital variables. We examine also various types of probabilistic relationships that may exist among variables. Independent evolution is an asymmetric relation, which is intended to encapsulate the notion of one process driving the other; $X$~is a driver of~$Y$ if $X$ evolves independently of the history of~$Y$. This concept arises in several places in the study of survival processes.

stat.ME↗

arXiv2016

Survival models and health sequences

Medical investigations focusing on patient survival often generate not only a failure time for each patient but also a sequence of measurements on patient health at annual or semi-annual check-ups while the patient remains alive. Such a sequence of random length accompanied by a survival time is called a survival process. Ordinarily robust health is associated with longer survival, so the two parts of a survival process cannot be assumed independent. This paper is concerned with a general technique---time reversal---for constructing statistical models for survival processes. A revival model is a regression model in the sense that it incorporates covariate and treatment effects into both the distribution of survival times and the joint distribution of health outcomes. It also allows individual health outcomes to be used clinically for predicting the subsequent survival time.

stat.ME↗

arXiv2015

Community detection for interaction networks

In many applications, it is common practice to obtain a network from interaction counts by thresholding each pairwise count at a prescribed value. Our analysis calls attention to the dependence of certain methods, notably Newman--Girvan modularity, on the choice of threshold. Essentially, the threshold either separates the network into clusters automatically, making the algorithm's job trivial, or erases all structure in the data, rendering clustering impossible. By fitting the original interaction counts as given, we show that minor modifications to classical statistical methods outperform the prevailing approaches for community detection from interaction datasets. We also introduce a new hidden Markov model for inferring community structures that vary over time. We demonstrate each of these features on three real datasets: the karate club dataset, voting data from the U.S.\ Senate (2001--2003), and temporal voting data for the U.S. Supreme Court (1990--2004).

cs.SI↗

arXiv2015

Atypical scaling behavior persists in real world interaction networks

Scale-free power law structure describes complex networks derived from a wide range of real world processes. The extensive literature focuses almost exclusively on networks with power law exponent strictly larger than 2, which can be explained by constant vertex growth and preferential attachment. The complementary scale-free behavior in the range between 1 and 2 has been mostly neglected as atypical because there is no known generating mechanism to explain how networks with this property form. However, empirical observations reveal that scaling in this range is an inherent feature of real world networks obtained from repeated interactions within a population, as in social, communication, and collaboration networks. A generative model explains the observed phenomenon through the realistic dynamics of constant edge growth and a positive feedback mechanism. Our investigation, therefore, yields a novel empirical observation grounded in a strong theoretical basis for its occurrence.

cs.SI↗