Real-Time Regression Analysis of Streaming Clustered Data With Possible Abnormal Data Batches

TL;DR

This paper introduces a renewable quadratic inference function (QIF) method for real-time analysis of streaming clustered data, enabling efficient incremental inference and abnormal data detection without raw historical data.

Contribution

It presents a novel renewable QIF algorithm for streaming data, with theoretical efficiency and a sequential goodness-of-fit test for abnormal data detection, integrated into Spark architecture.

Findings

RenewQIF achieves statistical and computational efficiency.

The method effectively detects abnormal data batches.

Simulation and real data demonstrate practical applicability.

Abstract

This paper develops an incremental learning algorithm based on quadratic inference function (QIF) to analyze streaming datasets with correlated outcomes such as longitudinal data and clustered data. We propose a renewable QIF (RenewQIF) method within a paradigm of renewable estimation and incremental inference, in which parameter estimates are recursively renewed with current data and summary statistics of historical data, but with no use of any historical subject-level raw data. We compare our renewable estimation method with both offline QIF and offline generalized estimating equations (GEE) approach that process the entire cumulative subject-level data, and show theoretically and numerically that our renewable procedure enjoys statistical and computational efficiency. We also propose an approach to diagnose the homogeneity assumption of regression coefficients via a sequential goodness-of-fit test as a screening procedure on occurrences of abnormal data batches. We implement the proposed methodology by expanding existing Spark's Lambda architecture for the operation of statistical inference and data quality diagnosis. We illustrate the proposed methodology by extensive simulation studies and an analysis of streaming car crash datasets from the National Automotive Sampling System-Crashworthiness Data System (NASS CDS).