A0358
Title: Matrix-valued factor model with time-varying main effects
Authors: Clifford Lam - London School of Economics and Political Science (United Kingdom)
Zetai Cen - London School of Economics and Political Science (United Kingdom) [presenting]
Abstract: A new proposal for the matrix-valued time-varying main effects factor model (MEFM) is presented. MEFM is a generalization to the traditional matrix-valued factor model (FM). Rigorous definitions of MEFM and its identifications are given, with estimators proposed for the time-varying grand mean, row and column main effects, and the row and column factor loading matrices for the common component. Rates of convergence for different estimators are spelt out, with asymptotic normality shown. The core rank estimator for the common component is also proposed, with the consistency of the estimators presented. A crucial test is proposed to test if FM is sufficient against the alternative that MEFM is necessary, whose power is demonstrated by various simulation settings. The accuracy of our estimators is also demonstrated numerically in extended simulation experiments. A set of NYC taxi traffic data is analyzed, and the proposed test suggests that MEFM is indeed necessary to analyze the data against a traditional FM.
