Resumen de Contributions to time series factor modeling: model averaging and bias correction
Guadalupe Bastos
Given the increasing availability of data and the evolution of computation, there is a growing body of theory and applications taking advantage of multivariate datasets. Following this line of research, we work with multivariate time series and employ the Factor Model. With the Factor Model we exploit the relation between the series, include their dynamic nature and keep a reduced number of parameters. The purpose of this dissertation is to improve the forecasts of high-dimensional vectors of time series when Factor Models are employed.
We work with forecast combination, motivated by the unsolved issues of selecting a number of common factors and selecting a ‘best’ model for each of them. Instead of employing a particular criterion for model selection, we estimate several specifications with alternative numbers of common factors and alternative models for them. Afterwards, we use the forecasts of these models and evaluate the performance of five easy to apply forecast combination techniques in an application to electricity prices of the Iberian and Italian markets.
On the other hand, it is not unusual to deal with bias in the coefficients of autoRegressive models, especially when the sample size has a small time dimension. We work with two techniques to correct the bias for highly persistent autoRegressive common factors. We employ simulations to show that the correction of the bias improves the forecasts for the variables of interest. We also observe this improvement using an illustrative dataset of Industrial Production Indexes for thirteen European countries.
