# 16-067/IV (2016-08-29)

Andre Lucas, VU University Amsterdam, the Netherlands; Anne Opschoor, VU University Amsterdam, the Netherlands; Julia Schaumburg, VU University Amsterdam, the Netherlands
generalized autoregressive score models, missing completely at random, Expectation-Maximization
JEL codes:
C52, C53

We show that two alternative perspectives on how to deal with missing data in the context of the score-driven time-varying parameter models of Creal, Koopman, Lucas (2013) and Harvey (2013) lead to precisely the same dynamic transition equations. As score-driven models encompass a wide variety of time-varying parameter models (including generalized autoregressive conditional volatility (GARCH) and duration (ACD) models), the results apply to a wide range of empirically relevant models as applied in economics and statistics.