Findes ikke på dansk
George Kapetanios
- 1 April 2015
- WORKING PAPER SERIES - No. 1773Details
- Abstract
- This paper assesses the forecasting performance of various variable reduction and variable selection methods. A small and a large set of wisely chosen variables are used in forecasting the industrial production growth for four Euro Area economies. The results indicate that the Automatic Leading Indicator (ALI) model performs well compared to other variable reduction methods in small datasets. However, Partial Least Squares and variable selection using heuristic optimisations of information criteria along with the ALI could be used in model averaging methodologies.
- JEL Code
- C11 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Bayesian Analysis: General
C32 : Mathematical and Quantitative Methods→Multiple or Simultaneous Equation Models, Multiple Variables→Time-Series Models, Dynamic Quantile Regressions, Dynamic Treatment Effect Models, Diffusion Processes
C52 : Mathematical and Quantitative Methods→Econometric Modeling→Model Evaluation, Validation, and Selection
- 13 May 2009
- WORKING PAPER SERIES - No. 1051Details
- Abstract
- Factor based forecasting has been at the forefront of developments in the macroeconometric forecasting literature in the recent past. Despite the flurry of activity in the area, a number of specification issues such as the choice of the number of factors in the forecasting regression, the benefits of combining factor-based forecasts and the choice of the dataset from which to extract the factors remain partly unaddressed. This paper provides a comprehensive empirical investigation of these issues using data for the euro area, the six largest euro area countries, and the UK.
- JEL Code
- C10 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→General
C15 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Statistical Simulation Methods: General
C53 : Mathematical and Quantitative Methods→Econometric Modeling→Forecasting and Prediction Methods, Simulation Methods
- 25 January 2008
- WORKING PAPER SERIES - No. 850Details
- Abstract
- Testing and estimating the rank of a matrix of estimated parameters is key in a large variety of econometric modelling scenarios. This paper describes general methods to test for and estimate the rank of a matrix, and provides details on a variety of modelling scenarios in the econometrics literature where such methods are required. Four different methods to test the true rank of a general matrix are described, as well as one method that can handle the case of a matrix subject to parameter constraints associated with defineteness structures. The technical requirements for the implementation of the tests of rank of a general matrix differ and hence there are merits to all of them that justify their use in applied work. Nonetheless, we review available evidence of their small sample properties in the context of different modelling scenarios where all, or some, are applicable.
- JEL Code
- C12 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Hypothesis Testing: General
C15 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Statistical Simulation Methods: General
C32 : Mathematical and Quantitative Methods→Multiple or Simultaneous Equation Models, Multiple Variables→Time-Series Models, Dynamic Quantile Regressions, Dynamic Treatment Effect Models, Diffusion Processes
- 5 November 2004
- WORKING PAPER SERIES - No. 402Details
- Abstract
- Standard measures of prices are often contaminated by transitory shocks. This has prompted economists to suggest the use of measures of underlying inflation to formulate monetary policy and assist in forecasting observed inflation. Recent work has concentrated on modelling large datasets using factor models. In this paper we estimate factors from datasets of disaggregated price indices for European countries. We then assess the forecasting ability of these factor estimates against other measures of underlying inflation built from more traditional methods. The power to forecast headline inflation over horizons of 12 to 18 months is adopted as a valid criterion to assess forecasting. Empirical results for the five largest euro area countries as well as for the euro area are presented.
- JEL Code
- E31 : Macroeconomics and Monetary Economics→Prices, Business Fluctuations, and Cycles→Price Level, Inflation, Deflation
C13 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Estimation: General
C32 : Mathematical and Quantitative Methods→Multiple or Simultaneous Equation Models, Multiple Variables→Time-Series Models, Dynamic Quantile Regressions, Dynamic Treatment Effect Models, Diffusion Processes
- 28 April 2004
- WORKING PAPER SERIES - No. 349Details
- Abstract
- The rank of the spectral density matrix conveys relevant information in a variety of statistical modelling scenarios. This note shows how to estimate the rank of the spectral density matrix at any given frequency. The method presented is valid for any hermitian positive definite matrix estimate that has a normal asymptotic distribution with a covariance matrix whose rank is known.
- JEL Code
- C12 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Hypothesis Testing: General
C32 : Mathematical and Quantitative Methods→Multiple or Simultaneous Equation Models, Multiple Variables→Time-Series Models, Dynamic Quantile Regressions, Dynamic Treatment Effect Models, Diffusion Processes
C52 : Mathematical and Quantitative Methods→Econometric Modeling→Model Evaluation, Validation, and Selection
- 1 April 2001
- WORKING PAPER SERIES - No. 62Details
- Abstract
- The rank of the spectral density matrix conveys relevant information in a variety of modelling scenarios. Phillips (1986) showed that a necessary condition for cointegration is that the spectral density matrix of the innovation sequence at frequency zero is of a reduced rank. In a recent paper Forni and Reichlin (1998) suggested the use of generalized dynamic factor model to explain the dynamics of a large set of macroeconomic series. Their method relied also on the computation of the rank of the spectral density matrix. This paper provides formal tests to estimate the rank of the spectral density matrix at any given frequency. The tests of rank at frequency zero are tests of the null of 'cointegration', complementary to those suggested by Phillips and Ouliaris (1988) which test the null of 'no cointegration'.
- JEL Code
- C12 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Hypothesis Testing: General
C15 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Statistical Simulation Methods: General
C32 : Mathematical and Quantitative Methods→Multiple or Simultaneous Equation Models, Multiple Variables→Time-Series Models, Dynamic Quantile Regressions, Dynamic Treatment Effect Models, Diffusion Processes
- 1 March 2001
- WORKING PAPER SERIES - No. 45Details
- Abstract
- The rank of the Hankel matrix, corresponding to a system transfer function, is equal to the order of its minimal state space realization. The computation of the rank of the Hankel matrix is complicated by the fact that its block elements are rarely given exactly but are estimated instead. In this paper, we propose new statistical tests to determine the rank of the Hankel matrix. We also provide a Monte Carlo study on the reliability of these tests compared to existing procedures.
- JEL Code
- C12 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Hypothesis Testing: General
C15 : Mathematical and Quantitative Methods→Econometric and Statistical Methods and Methodology: General→Statistical Simulation Methods: General
C32 : Mathematical and Quantitative Methods→Multiple or Simultaneous Equation Models, Multiple Variables→Time-Series Models, Dynamic Quantile Regressions, Dynamic Treatment Effect Models, Diffusion Processes