ESTIMATION AND ASYMPTOTIC THEORY FOR A NEW CLASS OF MIXTURE MODELS

In this paper a new model of mixture of distributions is proposed, where the mixing structure is determined by a smooth transition tree architecture. Models based on mixture of distributions are useful in order to approximate unknown conditional distributions of multivariate data. The tree structure yields a model that is simpler, and in some cases more interpretable, than previous proposals in the literature. Based on the Expectation-Maximization (EM) algorithm a quasi-maximum likelihood estimator is derived and its asymptotic properties are derived under mild regularity conditions. In addition, a specific-to-general model building strategy is proposed in order to avoid possible identification problems. Both the estimation procedure and the model building strategy are evaluated in a Monte Carlo experiment, which give strong support for the theory developed in small samples. The approximation capabilities of the model is also analyzed in a simulation experiment. Finally, two applications with real datasets are considered. KEYWORDS: Mixture models, smooth transition, EM algorithm, asymptotic properties, time series, conditional distribution.

[1]  Wai Keung Li,et al.  On a Mixture Autoregressive Conditional Heteroscedastic Model , 2001 .

[2]  C. Granger,et al.  Forecasting Volatility in Financial Markets: A Review , 2003 .

[3]  T. Bollerslev,et al.  Generalized autoregressive conditional heteroskedasticity , 1986 .

[4]  W. Li,et al.  On a mixture autoregressive model , 2000 .

[5]  C. Granger,et al.  Modelling Nonlinear Economic Relationships , 1995 .

[6]  Xiaotong Shen,et al.  Sieve extremum estimates for weakly dependent data , 1998 .

[7]  Marcelo C. Medeiros,et al.  A flexible coefficient smooth transition time series model , 2005, IEEE Transactions on Neural Networks.

[8]  Kurt Hornik,et al.  Universal approximation of an unknown mapping and its derivatives using multilayer feedforward networks , 1990, Neural Networks.

[9]  R. Adler,et al.  Non-Linear Models for Time Series Using Mixtures of Autoregressive Models , 2000 .

[10]  Timo Teräsvirta,et al.  Testing linearity against smooth transition autoregressive models , 1988 .

[11]  T. Teräsvirta.  Specification, Estimation, and Evaluation of Smooth Transition Autoregressive Models , 1994 .

[12]  A. Raftery,et al.  Modeling flat stretches, bursts, and outliers in time series using mixture transition distribution models , 1996 .

[13]  M. McAleer.  Automated Inference and Learning in Modelling Financial Volatility * , 2004 .

[14]  Peter F. Christoffersen.  Evaluating Interval Forecasts , 1998 .

[15]  Halbert White,et al.  On learning the derivatives of an unknown mapping with multilayer feedforward networks , 1992, Neural Networks.

[16]  Halbert White,et al.  Improved Rates and Asymptotic Normality for Nonparametric Neural Network Estimators , 1999, IEEE Trans. Inf. Theory.

[17]  Martin A. Tanner,et al.  Mixtures-of-experts of autoregressive time series: asymptotic normality and model specification , 2005, IEEE Transactions on Neural Networks.

[18]  P. Manimaran,et al.  Modelling Financial Time Series , 2006 .

[19]  Halbert White,et al.  Artificial neural networks: an econometric perspective ∗ , 1994 .

[20]  Howell Tong,et al.  Threshold autoregression, limit cycles and cyclical data- with discussion , 1980 .

[21]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[22]  R. Engle.  Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation , 1982 .

[23]  Marcelo C. Medeiros,et al.  Tree-structured smooth transition regression models , 2008, Comput. Stat. Data Anal..

[24]  Forecasting Volatility in Financial Markets , 2003 .

[25]  Robert A. Jacobs,et al.  Hierarchical Mixtures of Experts and the EM Algorithm , 1993, Neural Computation.

[26]  W. Härdle,et al.  A Review of Nonparametric Time Series Analysis , 1997 .

[27]  Steven J. Nowlan,et al.  Maximum Likelihood Competitive Learning , 1989, NIPS.

[28]  Gabriel Huerta,et al.  Time series modeling via hierarchical mixtures , 2003 .

[29]  H. Tong.  On a threshold model , 1978 .

[30]  W. Härdle.  Applied Nonparametric Regression , 1991 .

[31]  P. T. Szymanski,et al.  Adaptive mixtures of local experts are source coding solutions , 1993, IEEE International Conference on Neural Networks.

[32]  Jianqing Fan,et al.  Nonlinear Time Series : Nonparametric and Parametric Methods , 2005 .

[33]  H. Tong,et al.  ON ESTIMATING THRESHOLDS IN AUTOREGRESSIVE MODELS , 1986 .

[34]  M. Medeiros,et al.  Building Neural Network Models for Time Series: A Statistical Approach , 2002 .

[35]  Wenxin Jiang,et al.  On the identifiability of mixtures-of-experts , 1999, Neural Networks.

[36]  Alexandre X. Carvalho,et al.  Ergodicity and existence of moments for local mixtures of linear autoregressions , 2005 .

[37]  Timo Teräsvirta,et al.  Smooth transition autoregressive models - A survey of recent developments , 2000 .

[38]  Michael I. Jordan,et al.  Hierarchical Mixtures of Experts and the EM Algorithm , 1994, Neural Computation.

[39]  Halbert White,et al.  Estimation, inference, and specification analysis , 1996 .

[40]  Kurt Hornik,et al.  Stationary and Integrated Autoregressive Neural Network Processes , 2000, Neural Computation.

[41]  Stephen L Taylor,et al.  Modelling Financial Time Series , 1987 .

[42]  Geoffrey J. McLachlan,et al.  A Note on the Aitkin‐Rubin Approach to Hypothesis Testing in Mixture Models , 1987 .

[43]  Timo Teräsvirta,et al.  A SIMPLE VARIABLE SELECTION TECHNIQUE FOR NONLINEAR MODELS , 2001 .

[44]  Sally Wood,et al.  Bayesian mixture of splines for spatially adaptive nonparametric regression , 2002 .

[45]  Ashok N. Srivastava,et al.  Nonlinear gated experts for time series: discovering regimes and avoiding overfitting , 1995, Int. J. Neural Syst..

[46]  Donald B. Rubin,et al.  Max-imum Likelihood from Incomplete Data , 1972 .

[47]  David J. C. MacKay,et al.  Bayesian Interpolation , 1992, Neural Computation.