Building a theoretical and methodological framework for collaborative statistical inference and learning: multi-party and multiphase paradigms

点击翻译按钮获取中文摘要

Scientific data almost always undergo filtering, imputation, and other forms of preprocessing before they are analyzed. When such steps are taken, the data analysis becomes a collaborative endeavor by all parties involved in data collection, preprocessing, and inference. This research terms such settings as falling within the multiparty and multiphase paradigms for statistical inference and learning. These settings are rife with subtleties and pitfalls. Each party does not and often cannot have a perfect understanding of the entire phenomenon at hand; the final results will inevitably contain some combination of their judgments, and some preprocessing can irreversibly destroy information from the raw data. Building upon his previous theory and methods for dealing with uncongeniality with multiple imputation, the PI and his students aim to develop a set of statistical theory and methods to understand such problems and to provide better preprocessing, inferences, and learning. Their ultimate goals include providing methods for assessing the validity of such collaborative analyses, guidance on statistically-principled preprocessing, and a rich new theory of statistical learning and inference with multiple parties. The theoretical framework they develop can shed light on principles and methods for constructing more useful scientific databases, handling complex measurement processes, and analyzing massive datasets.With the dramatic increases in the size, diversity, and complexity of data available for scientific discoveries, medical advances, education reforms and evidence-based policy making, the entire enterprise of quantitative scientific inquiry has been presented with unprecedented challenges and opportunities. The vast majority of current inquiries are not made by a single individual or even a single team. In particular, the analysis of scientific data depends heavily on preprocessing in practice. Following data collection, raw data is typically transformed into a more easily handled form. Such transformations range from innocuous to highly destructive. When poorly executed, they can destroy huge scientific investments by rendering them useless for future analyses. These dangers are rising in importance as scientists and funding agencies emphasize the construction of scientific repositories and "big data". Despite its importance, preprocessing is poorly understood from a theoretical perspective. Even among statisticians, conventional wisdom and informal guidance are the norm. The PI and his students will work to close this gap, building a theory of preprocessing and collaborative inference. This theory aims to guide the construction of scientific repositories and the analysis of massive datasets generated by the latest technologies. It can also open the doors to greater collaboration and access to high-quality scientific data, broadening the scientific enterprise.