Predicting equity premium using dynamic model averaging. Does the state–space representation matter?

Dynamic model averaging (DMA) has become a very useful tool with regards to dealing with two important aspects of time-series analysis, namely, parameter instability and model uncertainty. An important component of DMA is the Kalman filter. It is used to filter out the latent time-varying regression coefficients of the predictive regression of interest, and produce the model predictive likelihood, which is needed to construct the probability of each model in the model set. To apply the Kalman filter, one must write the model of interest in linear state–space form. In this study, we demonstrate that the state–space representation has implications on out-of-sample prediction performance, and the degree of shrinkage. Using Monte Carlo simulations as well as financial data at different sampling frequencies, we document that the way in which the current literature tends to formulate the candidate time-varying parameter predictive regression in linear state–space form ignores empirical features that are often present in the data at hand, namely, predictor persistence and predictor endogeneity. We suggest a straightforward way to account for these features in the DMA setting. Results using the widely applied Goyal and Welch (2008) dataset document that modifying the DMA framework as we suggest has a bearing on equity premium point prediction performance from a statistical as well as an economic viewpoint.     

Misspecified models with dependent observations

General results are given in this paper which allow the development of a theory of estimation and inference for situations in which the model of a data-generating process has been misspecified. Observations may come from time-series, cross-section, panel, or experimental data. The nonlinear regression model is examined in some detail. Conditions are provided which ensure the consistency and asymptotic normality of the least-squares estimator with respect to the parameter vector of a weighted least-squares approximation to the underlying data-generating process. A specification-robust estimator of the asymptotic covariance matrix is given, allowing a proper treatment of inference in potentially misspecified models. The properties of the approximation and the covariance estimator are exploited to yield new tests for model specification.     

Three-stage least-squares estimation for a system of simultaneous,nonlinear, implicit equations

The article describes a nonlinear three-stage least-squares estimator for the parameters of a system of simultaneous, nonlinear, implicit equations; the method allows the estimation of these parameters subject to nonlinear parametric restrictions across equations. The estimator is shown to be strongly consistent, asymptotically normally distributed, and more efficient than the nonlinear two-stage least-squares estimator. Some practical implications of the regularity conditions used to obtain these results are discussed from the point of view of one whose interest is in applications, Also, computing methods using readily available nonlinear regression programs are described.     

Estimation of linear models with crossed-error structure

Sufficient conditions are presented under which the generalized least-squares estimator, with estimated covariance matrix, is unbiased for the parameters in the crossed-error model and has the same asymptotic distribution as the generalized least-squares estimator. The model permits the presence of independent variables that are constant over cross sections or time periods. The model does not require that the variance components associated with cross sections or time periods be positive.     

Volatility of Price Indices for Heterogeneous Goods with Applications to the Fine Art Market

Price indices for heterogeneous goods such as real estate or fine art constitute crucial information for institutional or private investors considering alternative investment decisions in times of financial markets turmoil. Classical mean‐variance analysis of alternative investments has been hampered by the lack of a systematic treatment of volatility in these markets. In this paper we propose a hedonic regression framework which explicitly defines an underlying stochastic process for the price index, allowing to treat the volatility parameter as the object of interest. The model can be estimated using maximum likelihood in combination with the Kalman filter. We derive theoretical properties of the volatility estimator and show that it outperforms the standard estimator. We show that extensions to allow for time‐varying volatility are straightforward using a local‐likelihood approach. In an application to a large data set of international blue chip artists, we show that volatility of the art market, although generally lower than that of financial markets, has risen after the financial crisis of 2008–09, but sharply decreased during the recent debt crisis. Copyright © 2013 John Wiley & Sons, Ltd.     

Model reference adaptive systems applied to regression analyses

The adaptive estimation procedure of model reference adaptive systems is modified and applied to linear models. In general the principle can be used for almost any time series model. Because of the recursive nature of the resulting estimator, it is computationally appealing, especially when a time series is considered as a flow of data. In addition, the estimator turns out to have certain statistical optimality properties.
In the linear regression setting, Ridge estimators turn out to constitute a subclass of the adaptive estimators considered, whereas for unknown measurement variance, the resulting estimators are related to J ames -S tkin type estimators, and have better properties than the latter. The estimator is shown to be strongly consistent and to converge in law to a normal variate under the standard assumptions of linear models. Further it is shown to be admissible and minimax in restricted parameter spaces. The connection between K alman filters and the classical least-squares estimator is also pointed out.     

On solving bias‐corrected non‐linear estimation equations with an application to the dynamic linear model

In a seminal paper, Mak, Journal of the Royal Statistical Society B, 55, 1993, 945, derived an efficient algorithm for solving non‐linear unbiased estimation equations. In this paper, we show that when Mak's algorithm is applied to biased estimation equations, it results in the estimates that would come from solving a bias‐corrected estimation equation, making it a consistent estimator if regularity conditions hold. In addition, the properties that Mak established for his algorithm also apply in the case of biased estimation equations but for estimates from the bias‐corrected equations. The marginal likelihood estimator is obtained when the approach is applied to both maximum likelihood and least squares estimation of the covariance matrix parameters in the general linear regression model. The new approach results in two new estimators when applied to the profile and marginal likelihood functions for estimating the lagged dependent variable coefficient in the dynamic linear regression model. Monte Carlo simulation results show the new approach leads to a better estimator when applied to the standard profile likelihood. It is therefore recommended for situations in which standard estimators are known to be biased.     

SCAD‐penalized quantile regression for high‐dimensional data analysis and variable selection

The present penalized quantile variable selection methods are only applicable to finite number of predictors or do not have oracle property associated with estimator. This technique is considered as an alternative to ordinary least squares regression in case of the outliers and the heavy‐tailed errors existing in linear models. The variable selection through quantile regression with diverging number of parameters is investigated in this paper. The convergence rate of estimator with smoothly clipped absolute deviation penalty function is also studied. Moreover, the oracle property with proper selection of tuning parameter for quantile regression under certain regularity conditions is also established. In addition, the rank correlation screening method is used to accommodate ultra‐high dimensional data settings. Monte Carlo simulations demonstrate finite performance of the proposed estimator. The results of real data reveal that this approach provides substantially more information as compared with ordinary least squares, conventional quantile regression, and quantile lasso.     

Robust estimation of ARIMA models

The paper discusses methods of estimating univariate ARIMA models with outliers. The approach calls for a state vector representation of a time-series model, on which we can then operate on using the Kalman filter. One of the additional advantages of Kalman filter operating on the state vector representation is that the method and code could easily be adapted to be applicable to the ARIMA model with missing observations. The paper investigates ways to calculate robust initial estimation of the parameters of the ARIMA model. The method proposed is based on the results obtained by R.D. Martin (1980).     

The nonlinear limited-information maximum- likelihood estimator and the modified nonlinear two-stage least-squares estimator

Several limited-information type estimators of the nonlinear simultaneous equation model are considered and their asymptotic covariance matrices are compared. Amemiya (1974) proposed the general class of nonlinear two-stage least-squares estimators. In this paper, its two specific members are considered and, in addition, the nonlinear limited-information maximum- likelihood estimator and the modified nonlinear two-stage least-squares estimator are proposed. Both are shown to be asymptotically more efficient than the nonlinear two-stage least-squares estimator, and the second has the advantage of being computationally simple.     

Reduced rank regression in cointegrated models

The coefficient matrix of a cointegrated first-order autoregression is estimated by reduced rank regression (RRR), depending on the larger canonical correlations and vectors of the first difference of the observed series and the lagged variables. In a suitable coordinate system the components of the least-squares (LS) estimator associated with the lagged nonstationary variables are of order 1/T, where T is the sample size, and are asymptotically functionals of a Brownian motion process; the components associated with the lagged stationary variables are of the order T^−1/2 and are asymptotically normal. The components of the RRR estimator associated with the stationary part are asymptotically the same as for the LS estimator. Some components of the RRR estimator associated with nonstationary regressors have zero error to order 1/T and the other components have a more concentrated distribution than the corresponding components of the LS estimator.     

Applying Kalman Filter on Solving Simultaneous Equations with Overidentifying Rank Restrictions: The Analysis of the Demand and Supply Model of Medium-size Scooter Market in Taiwan

Once the structure form of demand and supply is translated into areduced form, one can solve the reduced form with a state space modelof the Kalman filter method. This paper discusses an innovationrepresentation that links the structure form with the state space model.For the state space model, the recursive Expectation Maximization(EM) algorithm is used to estimate the parameters of a structure form.This research successfully applied the Kalman filter method to theestimation of the coefficients of simultaneous equations withoveridentifying rank restrictions. The empirical monthly data set camefrom the medium-size scooter market in Taiwan during 1987 to 1992period.     

Estimating Mann–Whitney‐type Causal Effects

Mann–Whitney‐type causal effects are generally applicable to outcome variables with a natural ordering, have been recommended for clinical trials because of their clinical relevance and interpretability and are particularly useful in analysing an ordinal composite outcome that combines an original primary outcome with death and possibly treatment discontinuation. In this article, we consider robust and efficient estimation of such causal effects in observational studies and clinical trials. For observational studies, we propose and compare several estimators: regression estimators based on an outcome regression (OR) model or a generalised probabilistic index (GPI) model, an inverse probability weighted estimator based on a propensity score model and two doubly robust (DR), locally efficient estimators. One of the DR estimators involves a propensity score model and an OR model, is consistent and asymptotically normal under the union of the two models and attains the semiparametric information bound when both models are correct. The other DR estimator has the same properties with the OR model replaced by a GPI model. For clinical trials, we extend an existing augmented estimator based on a GPI model and propose a new one based on an OR model. The methods are evaluated and compared in simulation experiments and applied to a clinical trial in cardiology and an observational study in obstetrics.     

One-step ahead adaptive D-optimal design on a finite design space is asymptotically optimal

We study the consistency of parameter estimators in adaptive designs generated by a one-step ahead D-optimal algorithm. We show that when the design space is finite, under mild conditions the least-squares estimator in a nonlinear regression model is strongly consistent and the information matrix evaluated at the current estimated value of the parameters strongly converges to the D-optimal matrix for the unknown true value of the parameters. A similar property is shown to hold for maximum-likelihood estimation in Bernoulli trials (dose–response experiments). Some examples are presented.     

Random coefficient state-space model: Estimation and performance in M3–M4 competitions

The random coefficient state-space model was first introduced by McKenzie and Gardner (2010). This model is a stochastic combination of simple and double exponential smoothing, a desirable feature for time-series forecasting. This paper provides a simple method to estimate the random coefficient state-space model parameters by exploiting the link between the model’s autocovariance and the Kalman filter. A simulation exercise shows that the proposed estimator has good finite-sample properties. This paper also evaluates the model’s forecasting performance in large-scale empirical applications, which is remarkable. Indeed, this model outperforms all competing (not-combined) benchmarks when using the yearly data from the M3 competition dataset. Furthermore, employing the yearly data from the M4 competition, it continues to beat its competitors, with a performance comparable to that of the Theta method. The predictive performance is assessed using both the MASE/sMAPE metrics and the Model Confidence Set procedure.     

On some properties of Markov chain Monte Carlo simulation methods based on the particle filter

Andrieu et al. (2010) prove that Markov chain Monte Carlo samplers still converge to the correct posterior distribution of the model parameters when the likelihood estimated by the particle filter (with a finite number of particles) is used instead of the likelihood. A critical issue for performance is the choice of the number of particles. We add the following contributions. First, we provide analytically derived, practical guidelines on the optimal number of particles to use. Second, we show that a fully adapted auxiliary particle filter is unbiased and can drastically decrease computing time compared to a standard particle filter. Third, we introduce a new estimator of the likelihood based on the output of the auxiliary particle filter and use the framework of Del Moral (2004) to provide a direct proof of the unbiasedness of the estimator. Fourth, we show that the results in the article apply more generally to Markov chain Monte Carlo sampling schemes with the likelihood estimated in an unbiased manner.     

Ridge regression revisited

In general ridge (GR) regression p ridge parameters have to be determined, whereas simple ridge regression requires the determination of only one parameter. In a recent textbook on linear regression, Jürgen Gross argues that this constitutes a major complication. However, as we show in this paper, the determination of these p parameters can fairly easily be done. Furthermore, we introduce a generalization of the GR estimator derived by Hemmerle and by Teekens and de Boer. This estimator, which is more conservative, performs better than the Hoerl and Kennard estimator in terms of a weighted quadratic loss criterion.     

Stochastic cointegration: estimation and inference

This paper considers the estimation of a stochastically cointegrating regression within the stochastic cointegration modelling framework introduced in McCabe et al. (Stochastic cointegration: testing, 2001). A stochastic cointegrating regression allows some or all of the variables to be conventionally or heteroscedastically integrated. This generalizes Hansen's (J. Econom. 54 (1992) 139) heteroscedastic cointegrating regression model, where the dependent variable is heteroscedastically integrated, but all the regressor variables are restricted to being conventionally integrated. In contrast to conventional and heteroscedastic cointegrating regression, ordinary least-squares (OLS) estimation is shown to be inconsistent, in general, in a stochastically cointegrating regression. As a solution, a new instrumental variables (IVs) estimator is proposed and is shown to be consistent. Under a suitable exogeneity assumption, standard asymptotic inference on the stochastic cointegrating vector can be carried out based on the IV estimator. The finite sample properties of the test statistics, including their robustness to the exogeneity assumption, are examined by simulation.     

Regression analysis with dichotomous regressors and misclassification

We discuss a regression model in which the regressors are dummy variables. The basic idea is that the observation units can be assigned to some well-defined combination of treatments, corresponding to the dummy variables. This assignment can not be done without some error, i.e. misclassification can play a role. This situation is analogous to regression with errors in variables. It is well-known that in these situations identification of the parameters is a prominent problem. We will first show that, in our case, the parameters are not identified by the first two moments but can be identified by the likelihood. Then we analyze two estimators. The first is a moment estimator involving moments up to the third order, and the second is a maximum likelihood estimator calculated with the help of the EM algorithm. Both estimators are evaluated on the basis of a small Monte Carlo experiment.