PhD defence Sofia Borodich Suarez

Keywords: Time-series forecasting, Panel data modelling, Big data clustering

"Essays in Econometrics: Nonlinear Panel Data, Time Series and Clustering"

This doctoral research focuses on improving how economists and data scientists analyze and interpret complex data. It develops new statistical methods that make it easier to draw reliable conclusions when studying people, firms, or countries over time, forecasting economic activity, and identifying patterns in large datasets.

The first part refines how researchers measure cause-and-effect relationships in models that include unobserved individual characteristics, reducing errors that often occur when samples are small or unbalanced. The second chapter extends this work to situations where decisions change over time, offering a clearer understanding of evolving behaviour. The third part presents a new method for forecasting industrial production by separating long-term trends from seasonal movements, which leads to more accurate predictions. The final chapter introduces an improved algorithm for grouping datasets that contain different types of variables.

Together, these contributions strengthen the accuracy, transparency, and reliability of modern economic analysis.

Click here for the live stream.