Forecasting Inflation: Factor Models and Factor Selection through Shrinkage by Nicholas Langel ’23, Statistics Colloquium, Wednesday, November 30, 1:10 – 1:50 pm, North Science Building 114, Wachenheim.
Abstract: Forecasting inflation is an important and difficult part of not only effective domestic monetary policy but also stable international economics. In this talk, I will introduce a multivariate statistical method called factor analysis as a possible way to model inflation. This model assumes that there are latent factors underlying the observed variables and hence reduces the dimension of them for easier interpretation and clearer understanding of their relationship. I will start from traditional factor analysis, discuss the model, the estimation and the interpretation. I will then introduce a relatively new approach to combine penalization methods with factor models which results in sparsity to further reduce dimensionality and improve interpretation.