Bayesian analysis of dynamic factor models

When forecasting macroeconomic variables, such as GDP-growth and inflation, a huge number of potentially informative time-series of varying quality are often available. Ignoring part of this information will necessarily lead to sub-optimal forecasts. On the other hand, utilizing all series in a statistically and computationally efficient manner is a formidable task. Within this project, so-called dynamic factor models are studied. Such models have become increasingly popular in finance and economics lately, and aim at summarizing the information content of a large number of variables in a few factors. In contrast to traditionally used factor analysis, dynamic factor models incorporate the serial dependence found in time-series data. The project aims at constructing statistically and computationally efficient forecasting methods as well as methods for model specification, with a particular focus on the number of factors to be used as well as detailed modelling of the time-series dynamics.

Publications

Karlsson, Sune, (2013), ‘Forecasting with Bayesian Vector Autoregressions’, ch. 15, p 791-897 in Elliot, G. and Timmernann, A., eds, Handbook of Economic Forecasting, vol 2B, Elsevier

Ding, Shutong, (2014), Model Choice in Bayesian VAR Models, PhD dissertation, Örebro University.

Karlsson, S. (2017). Corrigendum to “Bayesian reduced rank regression in econometrics” [J. Econometrics 75 (1996) 121–146]. Journal of Econometrics, 201 (1), 170-171.