Seminar Talk with Michael Pfarrhofer: Nowcasting with mixed-frequency data using Gaussian processes
We propose and discuss using Gaussian processes for mixed data sampling (MIDAS) regressions. This involves handling frequency mismatches, how to specify functional relationships between predictors and the dependent variable, and how to specify the error distribution of the predictive regressions. In a nowcasting and forecasting exercise, focusing on quarterly US output growth and inflation in the GDP deflator. We assess the trade-off between flexibility and parsimony along these three key modeling dimensions.
Michael Pfarrhofer is an assistant professor for economics on the Vienna University of Economics and Business.
Please register for this event at event@ihs.ac.at