Home > Research > Publications & Outputs > High-Dimensional Methods for Timely and Interpr...

Electronic data

  • 2023mosleyphd

    Final published version, 8.23 MB, PDF document

    Available under license: CC BY-NC-ND: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License

Text available via DOI:

View graph of relations

High-Dimensional Methods for Timely and Interpretable Economic Statistics

Research output: ThesisDoctoral Thesis

Publication date2023
Number of pages260
Awarding Institution
Thesis sponsors
  • Office for National Statistics
Award date15/08/2023
  • Lancaster University
<mark>Original language</mark>English


The need for a timely and detailed measure of the economy has become increasingly important in recent years due to sudden periods of economic growth and decline. It has become imperative for policymakers and practitioners to accurately capture short-term dynamics and trajectories as we emerge from shocks. However, relying on traditional survey-centric economic indicators presents challenges, as they often lack granularity and are published with significant delays. For this reason, National Statistics Institutes, such as the UK's Office for National Statistics, are undergoing a transformation by incorporating administrative and alternative data-sources as key components of their statistical frameworks. These data sources offer several benefits, including timeliness, cost-effectiveness, reliability, and comprehensive coverage of economic processes. While a sense of awareness for these new data-sources is being established, the statistical methodology to calibrate insights from these sources remains under-developed.

This thesis develops new methodology for producing timely and interpretable economic statistics. Specifically, it proposes novel techniques to address the crucial tasks of temporal disaggregation and nowcasting when dealing with large volumes of high-dimensional data. The frameworks developed in this research leverage the growing body of literature on sparsity-inducing regularisers, allowing for interpretable measures of relevant data-sources. Acknowledging the importance of reproducibility, the work presented is developed into R packages, and a wide range of case studies are extensively covered to showcase applicability of the proposed methodologies.