Research seminar: General-to-Specific Modelling and Indicator Saturation Methods with User-Specified Estimators and Models
- Genaro Succarat, BI
- 29 May 2018
- 11:45 - 12:45
- Auditoriet, SSB, Akersveien 26
"General-to-Specific Modelling and Indicator Saturation Methods with User-Specified Estimators and Models"
General-to-Specific (GETS) modelling provides a comprehensive, systematic and cumulative approach to modelling ideally suited for forecasting and policy analysis, whereas Indicator Saturation (IS) is a powerful and flexible approach to the detection and estimation of structural break (e.g. location shifts). The implementation of these methods, however, puts a large programming-burden on the user, and may require substantial computing power. We develop a framework for user-specified estimators and models, and provide flexible and computationally efficient free software that enables parallel computing on large clusters. The GETS and IS implementations are via the $R$ package $gets$, but user-specified estimators and models can - in fact - be implemented in an external language (e.g. C/C++, Fortran, Python, STATA, EViews, MATLAB, etc.).
NOTE: This is work in progress. Background reading: F. Pretis, J. Reade and G. Sucarrat, "Automated General-to-Specifi c (GETS) Modelling of the Mean and Variance of Regressions, and Indicator Saturation Methods for Outliers and Structural Breaks", J. Statistical Software (forthcoming)