Notes on ''On factor models with random missing: EM estimation, inference, and cross validation''

0. Background

1. Introduction

• Factor model in balanced panel has been thoroughly investigated.
• How to handle the missing data problem in factor models ?
  • the expectation–maximization (EM) algorithm
  • the Kalman filter (KF)
• There is no formal study of the asymptotic properties for the EM estimators of the factors and factor loadings for the PC estimation with missing observations

2. Large dimensional factor models with random missing

2.1. EM estimation

3. Determining the number of factors via cross validation

4. Monte Carlo simulations

5. Empirical application: Forecasting macroeconomic variables