The PRObability FORecasting (PROFOR) toolbox is part of a research project on “Probability Forecasting with Macro Variables” at Warwick Business School, University of Warwick. The main purpose of the project is to develop, extend and apply procedures to undertake probability forecasting with macro variables. In particular, the toolbox incorporates forecasting models and methods for combining forecasts, together with forecast evaluation metrics. The initial "demo" version provides Bayesian and frequentist estimation for a variety of time series models, including constant parameter autoregressive models, with time varying parameter and stochastic volatility variants. Subsequent versions will incorporate methods for estimation of mixed frequency models, plus an enhanced capacity for modelling non-linear and non-Gaussian processes. The demo toolbox runs in MATLAB R2014a and R2014b.
The demo version of the PROFOR toolbox contains the steps required for replication of the main results in two initial research papers: an application forecasting inflation; and a literature review on probability forecasting with macro variables. By facilitating code and data sharing, the aim is to make the PROFOR toolbox the starting point for graduate researchers and others wishing to learn advanced probability forecasting techniques.
The PROFOR toolbox, intended as a research tool, provides a steam of essential data and techniques useful for researchers wishing to develop a practical forecasting system. Adaptation to this end requires further modifications to the demo version.
The PROFOR toolbox was developed with financial support from Norges Bank, the Bank of England and Warwick Business School.
Demo version PROFOR 1.0
Bayesian and frequentist forecasting with (vector) autoregressions, time varying parameter models, and stochastic volatility variants. Download here: https://github.com/ProforTeam/profor