Time-varying combinations of predictive densities using nonlinear filtering

We propose a Bayesian combination approach for multivariate predictive densities which relies upon a distributional state space representation of the combination weights. Several specifications of multivariate time-varying weights are introduced with a particular focus on weight dynamics driven by the past performance of the predictive densities and the use of learning mechanisms. In the proposed approach the model set can be incomplete, meaning that all models can be individually misspecified. A Sequential Monte Carlo method is proposed to approximate the filtering and predictive densities. The combination approach is assessed using statistical and utility-based performance measures for evaluating density forecasts of simulated data, US macroeconomic time series and surveys of stock market prices. Simulation results indicate that, for a set of linear autoregressive models, the combination strategy is successful in selecting, with probability close to one, the true model when the model set is complete and it is able to detect parameter instability when the model set includes the true model that has generated subsamples of data. Also, substantial uncertainty appears in the weights when predictors are similar; residual uncertainty reduces when the model set is complete; and learning reduces this uncertainty. For the macro series we find that incompleteness of the models is relatively large in the 1970’s, the beginning of the 1980’s and during the recent financial crisis, and lower during the Great Moderation; the predicted probabilities of recession accurately compare with the NBER business cycle dating; model weights have substantial uncertainty attached. With respect to returns of the S&P 500 series, we find that an investment strategy using a combination of predictions from professional forecasters and from a white noise model puts more weight on the white noise model in the beginning of the 1990’s and switches to giving more weight to the professional forecasts over time. Information on the complete predictive distribution and not just on some moments turns out to be very important, above all during turbulent times such as the recent financial crisis. More generally, the proposed distributional state space representation offers great flexibility in combining densities.     

Estimating the LQAC model with I(2) variables

This paper derives a method for estimating and testing the Linear Quadratic Adjustment Cost (LQAC) model when the target variable and some of the forcing variables follow I(2) processes. Based on a forward-looking error-correction formulation of the model it is shown how to obtain strongly consistent estimates of the structural parameters from both a linear and a non-linear cointegrating regression where first-differences of the I(2) variables are included as regressors (multicointegration). Further, based on the estimated parameter values, it is shown how to test and evaluate the LQAC model using a VAR approach. A simple easy interpretable metric for measuring the model fit is suggested. In an empirical application using UK money demand data, the non-linear multicointegrating regression delivers an economically plausible estimate of the adjustment cost parameter. However, the restrictions implied by the exact LQAC model under rational expectations are strongly rejected and the metric for model fit indicates a substantial noise component in the model. Copyright © 1999 John Wiley & Sons, Ltd.     

基于改进粒子群算法的SVR参数优化选择的研究

支持向量回归机(SVR)模型的拟合精度和泛化能力取决于其相关参数的选取,由于在参数的选择范围内可选择的数量是无穷的,在多个参数中盲目搜索最优参数是需要极大的时间代价,并且很难逼近最优。因此提出了基于改进粒子群算法的SVR参数优化选择方法。仿真结果表明:该改进粒子群算法优化SVR参数方法可行、有效,由此得到的SVR模型具有更好的学习精度和推广能力。     

Estimation of spatial autoregressive panel data models with fixed effects

This paper establishes asymptotic properties of quasi-maximum likelihood estimators for SAR panel data models with fixed effects and SAR disturbances. A direct approach is to estimate all the parameters including the fixed effects. Because of the incidental parameter problem, some parameter estimators may be inconsistent or their distributions are not properly centered. We propose an alternative estimation method based on transformation which yields consistent estimators with properly centered distributions. For the model with individual effects only, the direct approach does not yield a consistent estimator of the variance parameter unless T is large, but the estimators for other common parameters are the same as those of the transformation approach. We also consider the estimation of the model with both individual and time effects.     

Inference on Reliability in Two-parameter Exponential Stress–strength Model

The problem of hypothesis testing and interval estimation of the reliability parameter in a stress–strength model involving two-parameter exponential distributions is considered. Test and interval estimation procedures based on the generalized variable approach are given. Statistical properties of the generalized variable approach and an asymptotic method are evaluated by Monte Carlo simulation. Simulation studies show that the proposed generalized variable approach is satisfactory for practical applications while the asymptotic approach is not satisfactory even for large samples. The results are illustrated using simulated data.     

Structural VAR and financial networks: A minimum distance approach to spatial modeling

In this paper, I interpret a time series spatial model (T-SAR) as a constrained structural vector autoregressive (SVAR) model. Based on these restrictions, I propose a minimum distance approach to estimate the (row-standardized) network matrix and the overall network influence parameter of the T-SAR from the SVAR estimates. I also develop a Wald-type test to assess the distance between these two models. To implement the methodology, I discuss machine learning methods as one possible identification strategy of SVAR models. Finally, I illustrate the methodology through an application to volatility spillovers across major stock markets using daily realized volatility data for 2004–2018.     

Monitoring mortality: a state-space approach

"A state-space model is developed which provides estimates of decrements in a dynamic environment. The model integrates the actual unfolding experience and a priori or Bayesian views of the rates. The estimates of present rates and predicted future rates are continually updated and associated standard errors have simple expressions. The model is described and applied in the context of mortality estimation but it should prove useful in other actuarial applications. The approach is particularly suitable for dynamic environments where data are scarce and updated parameter estimates are required on a regular basis. To illustrate the method it is used to monitor the unfolding mortality experience of the retired lives under an actual pension plan."     

Time varying VARs with inequality restrictions

In many applications involving time-varying parameter VARs, it is desirable to restrict the VAR coefficients at each point in time to be non-explosive. This is an example of a problem where inequality restrictions are imposed on states in a state space model. In this paper, we describe how existing MCMC algorithms for imposing such inequality restrictions can work poorly (or not at all) and suggest alternative algorithms which exhibit better performance. Furthermore, we show that previous algorithms involve an approximation relating to a key prior integrating constant. Our algorithms are exact, not involving this approximation. In an application involving a commonly used U.S. data set, we present evidence that the algorithms proposed in this paper work well.     

Learning in a multilateral bargaining experiment

This paper analyzes data from an investigation of a majoritarian bargaining experiment. A learning model is proposed to account for the evolution of play in this experiment. It is also suggested that an adjustment must be made to account for the panel structure of the data. Such adjustments have been used in other fields and are known to be important as unadjusted standard errors may be severely biased downward. These results indicate that this adjustment also has an important effect in this application. Furthermore, an efficient estimator that takes into account heterogeneity across players is proposed. A unique learning model to account for the paths of play under two different amendment rules cannot be rejected with the standard estimator with adjusted standard errors, however it can be rejected using the efficient estimator. The data and the estimated learning model suggest that after proposing “fair” divisions, subjects adapt and their proposals change rapidly in the treatment where uneven proposals are almost always accepted. Their beliefs in the estimated learning model are influenced by more than just the most recent outcomes.     

Transition time paths for overlapping-generations models

This paper examines the transition stages between steady states for an overlapping-generations growth model. Our procedure is based on eigenvalues and eigenvectors. For marginal parameter changes, we can generate exact ‘multipliers’ for the responses of state variables in each time period. Our solution technique is direct (rather than iterative), it yields an intermediate-stage test of stability, and it clearly reveals the need to interpret initial conditions carefully. We work several illustrative examples numerically.     

Supply chain planning of vaccine and pharmaceutical clusters under uncertainty: The case of COVID-19

As an abrupt epidemic occurs, healthcare systems are shocked by the surge in the number of susceptible patients' demands, and decision-makers mostly rely on their frame of reference for urgent decision-making. Many reports have declared the COVID-19 impediments to trading and global economic growth. This study aims to provide a mathematical model to support pharmaceutical supply chain planning during the COVID-19 epidemic. Additionally, it aims to offer new insights into hospital supply chain problems by unifying cold and non-cold chains and considering a wide range of pharmaceuticals and vaccines. This approach is unprecedented and includes an analysis of various pharmaceutical features such as temperature, shelf life, priority, and clustering. To propose a model for planning the pharmaceutical supply chains, a mixed-integer linear programming (MILP) model is used for a four-echelon supply chain design. This model aims to minimize the costs involved in the pharmaceutical supply chain by maintaining an acceptable service level. Also, this paper considers uncertainty as an intrinsic part of the problem and addresses it through the wait-and-see method. Furthermore, an unexplored unsupervised learning method in the realm of supply chain planning has been used to cluster the pharmaceuticals and the vaccines and its merits and drawbacks are proposed. A case of Tehran hospitals with real data has been used to show the model's capabilities, as well. Based on the obtained results, the proposed approach is able to reach the optimum service level in the COVID conditions while maintaining a reduced cost. The experiment illustrates that the hospitals' adjacency and emergency orders alleviated the service level significantly. The proposed MILP model has proven to be efficient in providing a practical intuition for decision-makers. The clustering technique reduced the size of the problem and the time required to solve the model considerably.     

Semiparametric Bayesian inference in smooth coefficient models

We describe procedures for Bayesian estimation and testing in cross-sectional, panel data and nonlinear smooth coefficient models. The smooth coefficient model is a generalization of the partially linear or additive model wherein coefficients on linear explanatory variables are treated as unknown functions of an observable covariate. In the approach we describe, points on the regression lines are regarded as unknown parameters and priors are placed on differences between adjacent points to introduce the potential for smoothing the curves. The algorithms we describe are quite simple to implement—for example, estimation, testing and smoothing parameter selection can be carried out analytically in the cross-sectional smooth coefficient model.     

Effectiveness of Q-learning as a tool for calibrating agent-based supply network models

This paper examines effectiveness of Q-learning as a tool for specifying agent attributes and behaviours in agent-based supply network models. Agent-based modelling (ABM) has been increasingly employed to study supply chain and supply network problems. A challenging task in building agent-based supply network models is to properly specify agent attributes and behaviours. Machine learning techniques, such as Q-learning, can be a useful tool for this purpose. Q-learning is a reinforcement learning technique that has been shown to be an effective adaptation and searching mechanism in distributed settings. In this study, Q-learning is employed by supply network agents to search for ‘optimal’ values for a parameter in their operating policies simultaneously and independently. Methods are designed to identify the ‘optimal’ parameter values against which effectiveness of the learning is evaluated. Robustness of the learning's effectiveness is also examined through consideration of different model settings and scenarios. Results show that Q-learning is very effective in finding the ‘optimal’ parameter values in all model settings and scenarios considered.     

Partial Minimax Estimation in Regression Analysis

The general minimax estimator of the linear regression model is applicable when the whole parameter vector is restricted to an ellipsoid. In many applications, however, it is more realistic to assume that only a part of the parameter set is constrained. For this case an alternative minimax approach is developed.     

Multilevel theory and the underspecification of multilevel models

This article emphasises the importance of the adequate specification of models of multilevel analysis in accordance with multilevel theories. Until recent times, multilevel theories tried only to explain the direct effect of group characteristics on an individual's characteristic. It seems to be more suited to adopt a more general theoretical approach, in which it is assumed that group characteristics affects individual processes. There a treshold effect and a process effect have to be distinguished. The propositions result in a model specification within the random coefficient model of multilevel analysis. The theory and model recommended are illustrated by means of data of Dar and Resh's (1986) study into social learning environment.     

Understanding Organizational Learning Capability

This paper presents data on how learning takes place in four organizations, what gets learned, and the factors and processes that facilitate or impede learning. Seven orientations for describing organizational learning capability and understanding learning styles are identified. Each of these orientations is conceived as a bi-polar continuum that reflect learning processes. Knowledge source is defined as the extent to which an organization prefers to develop new knowledge internally versus the extent to which it is more likely to seek inspiration in ideas developed externally. Product-process focus refers to a preference for the accumulation of knowledge related to product and service outcomes versus a preference to invest in knowledge about basic processes that support products. Documentation mode refers to attitudes as to what constitutes knowledge and to the repositories of knowledge that are supported. Dissemination mode pertains to the difference between establishing an atmosphere in which learning evolves and one in which a more structured, controlled approach is taken to induce learning. Learning focus has to do with whether learning is concentrated on methods and tools to improve what is already being done versus testing the assumptions underlying what is being done. Value-chain focus indicates which functional, core competencies are valued and supported. Skill development focus involves the orientation toward individual versus collective learning. Organizational learning may be increased by building on existing capabilities or developing new ones. the latter involves a change in culture, the former involves improving current capabilities. Organizations can enhance their learning capability through either approach.     

An engineering approach to LSE modelling of experience curves in the electricity supply industry

This review is a supplement to the paper by Sharp and Price (1990) and should be regarded as an alternative engineering approach to the modelling and forecasting of experience, or learning, curves. It highlights the problems associated with accurately defining a model to time series that show a combination of a continuous trend and a cyclical component, as detected by the authors in the Sharp and Price data. The authors give a number of alternative perspectives of the same time series, in this case average thermal efficiency data from the U.K. electricity supply industry, with the corresponding conclusions associated with each approach. Particular attention is drawn to the use of the “time constant learning curve” quoted by Sharp and Price which the authors show is a reasonable predictor of the average thermal efficiency. However, a tremendous improvement results from selecting the “ripple” model as a thermal efficiency predictor.     

基于支持向量机的鸡蛋供应链中价格预警研究

鸡蛋价格的大幅波动对鸡蛋供应链中生产者、经营者和消费者产生严重的影v向,因此鸡蛋供应链中的价格预警问题亟待研究。支持向量机是一种建立在统计学习理论基础上的机器学习方法,能够较好地解决小样本、非线性和局部极小点等问题。文中基于支持向量机方法建立鸡蛋价格预警模型,运用libsvm软件对样本集进行参数寻优、训练和测试。结果表明模型预警结果与实际情况相符,模型可以帮助鸡蛋供应链中的各主体在采购、生产、仓储、销售等方面更好地决策。     

Differences Between Classical and Bayesian Estimates for Mixed Logit Models: A Replication Study

The mixed logit model is widely used in applied econometrics. Researchers typically rely on the free choice between the classical and Bayesian estimation approach. However, empirical evidence of the similarity of their parameter estimates is sparse. The presumed similarity is mainly based on one empirical study that analyzes a single dataset (Huber J, Train KE. 2001. On the similarity of classical and Bayesian estimates of individual mean partworths. Marketing Letters 12 (3): 259–269). Our replication study offers a generalization of their results by comparing classical and Bayesian parameter estimates from six additional datasets and specifically for panel versus cross‐sectional data. In general, our results suggest that the two methods provide similar results, with less similarity for cross‐sectional data than for panel data. Copyright © 2016 John Wiley & Sons, Ltd.