Estimating Sparse Peer-Effect Networks with Low-Rank Structure
-
SeriesResearch Master Defense
-
Speaker
-
LocationErasmus University ET-48
Rotterdam -
Date and time
July 02, 2026
10:30 - 12:30
This thesis studies the estimation of peer-effect networks in panel data models, where each agent’s outcome may depend on the outcomes of other connected agents. Instead of treating the network matrix as fully unrestricted, the thesis imposes a low-rank structure that represents each link as the product of a receiver-specific and a transmitter-specific component. This structure reduces dimensionality and allows the model to capture heterogeneous exposure and influence patterns. The thesis develops and evaluates a penalised iterative estimator for recovering the latent network and compares its performance with an existing sparse-network estimator based on adaptive elastic net methods.