Simula@BI: Identifying dominant units using graphical models in panel time series data

We propose an extension of the graphical model to estimate the covariance matrix using LASSO estimators based on Meinshausen and Buehlmann (2006). Our extension allows the use of time dependent data and aims to identify dominant units in a network.

In detail, we propose to loop through the columns of a data matrix which represents the cross-sectional units. Within each  loop we obtain a selection of relevant regressors, which inform about the dependence structure of the data.

We carry out a Monte Carlo simulation to show that our estimator correctly identifies the  dominant units. We illustrate our method by applying it to a dataset of house prices in England.

• Read more about the speaker
• Visit our website to read more about the research center Simula@BI