Statistical Design and Inference in the Era of Ubiquitous Data

Statistical design has undergone a profound transformation in the age of ubiquitous data. Historically, the central challenge was how to extract reliable insight from carefully crafted and costly experiments. Classic designs such as randomized trials, stratified surveys, or factorial experiments were developed to maximize precision while minimizing error in resource-constrained settings. Today, by contrast, we are surrounded by vast streams of passively collected information: digital traces from online behavior, administrative records, wearable sensors, mobile health technologies, and other large-scale data systems. This abundance creates the illusion that statistical design might be less relevant. Yet in practice, the abundant but heterogeneous data elements often raise new and often more difficult questions. When data are opportunistically generated rather than deliberately designed, are they fit for purpose? What sources of error are introduced by gaps in coverage, hidden biases in collection processes, or measurement error from imperfect instruments? How should we design sampling or analytic strategies when the underlying distributions are unknown or shifting? These questions are central to modern practice in fields as varied as public health, economics, biomedical research, and the social sciences. Design principles developed for controlled settings remain crucial in this new landscape. Approaches such as blocking, clustering, factorial structures, and poststratification continue to provide the scaffolding for valid causal inference, even when researchers rely on hybrid or observational data. In parallel, frameworks originally built to assess and mitigate error are now being adapted to contexts where the quality of data is uneven and the boundaries of the population of interest are uncertain. Increasingly, passive data streams are supplemented with intentional design interventions to improve representativeness and extend inference beyond those who generate digital traces. But these strategies raise their own challenges: How do we characterize the population distribution in the first place? How do we determine whether the supplemented data are sufficient to recover parameters of interest? These questions highlight the evolving interplay between opportunistic and deliberate design. Biomedical and digital health research illustrate these issues especially vividly. Here, high-volume multimodal data, such as that of genomics, wearables, and clinical records, must be integrated with carefully designed experiments to generate actionable insights. The combination of passive and active data streams underscores both the promise and the pitfalls of abundant data. While offering unprecedented opportunities for discovery, they demand new strategies for assessing generalizability, mitigating error, and ensuring that results translate into practice. These perspectives highlight that statistical design is not being eclipsed, but rather expanding its scope. The enduring principles of design of experiments, sampling, and inference are being reinterpreted to anchor decision-making in a rapidly evolving data ecosystem. This session will show how the art of asking the right statistical question remains as important as ever. Far from diminishing the role of statisticians, the age of ubiquitous data underscores the need for thoughtful, principled design to transform abundance into knowledge.