Optimal error and GMDH predictors: A comparison with some statistical techniques

This paper presents some results obtained in time series forecasting using two nonstandard approaches and compares them with those obtained by usual statistical techniques. In particular, a new method based on recent results of the General Theory of Optimal Algorithm is considered. This method may be useful when no reliable statistical hypotheses can be made or when a limited number of observations is available. Moreover, a nonlinear modelling technique based on Group Method of Data Handling (GMDH) is also considered to derive forecasts. The well-known Wolf Sunspot Numbers and Annual Canadian Lynx Trappings series are analyzed; the Optimal Error Predictor is also applied to a recently published demographic series on Australian Births. The reported results show that the Optimal Error and GMDH predictors provide accurate one step ahead forecasts with respect to those obtained by some linear and nonlinear statistical models. Furthermore, the Optimal Error Predictor shows very good performances in multistep forecasting.     

GMDH网络在短期负荷预测中的应用

文章提出基于自组织方法的GMDH(Group Method of Data Handling)型神经网络并将它应用于短期负荷预测.与一般的前馈神经网络不同,GMDH网络的结构确定于训练过程之中,因而可大大提高神经网络性能.它能充分、合理地利用数据,自动进行变量组合,筛选及判断从而得到合适的模型,特别适用于数据预测.将这种用自组织方法所构成的GMDH型神经网络应用于广西某地区电力局的短期负荷预测,采用Matlab6.5进行仿真实验,证明其在短期负荷预测方面有很好的应用前景.     

GMDH网络在短期负荷预测中的应用

文章提出基于自组织方法的GMDH（Group Method of Data Halldlin亩型神经网络并将它应用于短期负荷预测。与一般的前馈神经网络不同,GMDH网络的结构确定于训练过程之中,因而可大大提高神经网络性能。它能充分、合理地利用数据,自动进行变量组合,筛选及判断从而得到合适的模型,特别适用于数据预测。将这种用自组织方法所构成的GMDH型神经网络应用于广西某地区电力局的短期负荷预测,采用Matlab6．5进行仿真实验,证明其在短期负荷预测方面有很好的应用前景。     

A non-linear filtering approach to stochastic volatility models with an application to daily stock returns

This paper develops a new model for the analysis of stochastic volatility (SV) models. Since volatility is a latent variable in SV models, it is difficult to evaluate the exact likelihood. In this paper, a non-linear filter which yields the exact likelihood of SV models is employed. Solving a series of integrals in this filter by piecewise linear approximations with randomly chosen nodes produces the likelihood, which is maximized to obtain estimates of the SV parameters. A smoothing algorithm for volatility estimation is also constructed. Monte Carlo experiments show that the method performs well with respect to both parameter estimates and volatility estimates. We illustrate our model by analysing daily stock returns on the Tokyo Stock Exchange. Since the method can be applied to more general models, the SV model is extended so that several characteristics of daily stock returns are allowed, and this more general model is also estimated. Copyright © 1999 John Wiley & Sons, Ltd.     

On the equivalence of location choice models: Conditional logit, nested logit and Poisson

It is well understood that the two most popular empirical models of location choice - conditional logit and Poisson - return identical coefficient estimates when the regressors are not individual specific. We show that these two models differ starkly in terms of their implied predictions. The conditional logit model represents a zero-sum world, in which one region’s gain is the other regions’ loss. In contrast, the Poisson model implies a positive-sum economy, in which one region’s gain is no other region’s loss. We also show that all intermediate cases can be represented as a nested logit model with a single outside option. The nested logit turns out to be a linear combination of the conditional logit and Poisson models. Conditional logit and Poisson elasticities mark the polar cases and can therefore serve as boundary values in applied research.     

Generalized method of moments and inverse control

This paper presents a generalized method of moments algorithm for estimating the structural parameters of a macroeconomic model subject to the restriction that the coefficients of the monetary policy rule minimize the central bank's expected loss function. The algorithm combines least-squares normal equations with moment restrictions derived from the first-order necessary conditions of the auxiliary optimization. We assess the performance of the algorithm with Monte Carlo simulations using three increasingly complex models. We find that imposing the optimizing restrictions when they are true improves estimation accuracy and that imposing those restrictions when they are false biases estimates of some of the structural parameters but not of the policy-rule coefficients.     

Mixed Hidden Markov Models for Longitudinal Data: An Overview

In this paper we review statistical methods which combine hidden Markov models (HMMs) and random effects models in a longitudinal setting, leading to the class of so‐called mixed HMMs. This class of models has several interesting features. It deals with the dependence of a response variable on covariates, serial dependence, and unobserved heterogeneity in an HMM framework. It exploits the properties of HMMs, such as the relatively simple dependence structure and the efficient computational procedure, and allows one to handle a variety of real‐world time‐dependent data. We give details of the Expectation‐Maximization algorithm for computing the maximum likelihood estimates of model parameters and we illustrate the method with two real applications describing the relationship between patent counts and research and development expenditures, and between stock and market returns via the Capital Asset Pricing Model.     

Adjusting for differential misclassification in multilevel models: the relationship between child exposure to smoke and cognitive development

Misclassification is found in many of the variables used in social sciences and, in practice, tends to be ignored in statistical analyses, and this can lead to biased results. This paper shows how to correct for differential misclassification in multilevel models and illustrates the extent to which this changes fixed and random parameter estimates. Reliability studies on self-reported behaviour of pregnant women suggest that there may be differential misclassification related to smoking and, thus, to child exposure to smoke. Models are applied to the Millennium Cohort Study data. The response variable is the child cognitive development assessed by the British Ability Scales at 3 years of age and explanatory variables are child exposure to smoke and family income. The proposed method allows a correction for misclassification when the specificity and sensitivity are known, and the assessment of potential biases occurring in the multilevel model parameter estimates if a validation data sample is not available, which is often the case.     

Estimating latent class model parameters for filter questions with skip patterns

Filter questions with skip patterns have been widely used in survey research, and latent class models (LCM) are often used to analyze this type of categorical data. The LCM parameters are usually estimated by means of an EM (expectation maximization) algorithm. When the pattern is present, the non-response of the skip pattern cannot be treated as random missingness. We thus propose a modified algorithm to estimate the latent class parameters when non-response is present, and the approach is attractive for two reasons. First, the latent class model with the algorithm is very flexible in the sense that it can model the association of variables with the skip patterns under study. Secondly, the algorithm can be easily implemented using any computer language. An empirical example is used to demonstrate the usefulness of the algorithm. The algorithm may also be flexibly generalized to more complex surveys, for example, polytomous responses.     

A Note on the Identification of Dynamic Economic Models with Generalized Shock Processes

Dynamic stochastic general equilibrium (DSGE) models with generalized shock processes, such as shock processes which follow a vector autoregression (VAR), have been an active area of research in recent years. Unfortunately, the structural parameters governing DSGE models are not identified when the driving process behind the model follows an unrestricted VAR. This finding implies that parameter estimates derived from recent attempts to estimate DSGE models with generalized driving processes should be treated with caution, and that there always exists a tradeoff between identification and the risk of model misspecification. However, these results also make it easier to address the issue of model misspecification by making it computationally easier to check the validity of cross‐equation restrictions.     

The Endogenous Economist:

A bstract . In the social sciences there exist links of communication and influence between the scientists and the actors of the modelled social systems. Social scientists are therefore endogenous parts of the systems which they study. Theories and models shouid be self-referential. The treatment of information in the rational choice models of the new classical theory, with its core assumption of rational expectations , has stopped short of extending the concept of endogeneity to the economists themselves. Economic theory proceeds on the assumption that it yields the true model and that the actors in the system share this same model even though economists have not yet agreed on this unique model. Moreover, it assumes that economists' disputes about the true model do not prompt the actors of the system to take positions and opt for one or the other of the discussed models. The use of a unique model perspective to interpret the response of financial markets to monetary announcements leads to inconclusive interpretations of the evidence. These findings are better interpreted as the result of multi model competition that has spilled over from academic science to the financial markets themselves.     

Spline-based quasi-likelihood estimation of mixed Poisson regression with single-index models

We consider spline-based quasi-likelihood estimation for mixed Poisson regression with single-index models. The unknown smooth function is approximated by B-splines, and a modified Fisher scoring algorithm is employed to compute the estimates. The spline estimate of the nonparametric component is shown to achieve the optimal rate of convergence, and the asymptotic normality of the regression parameter estimates is still valid even if the variance function is misspecified. The semiparametric efficiency of the model can be established if the variance function is correctly specified. The variance of the regression parameter estimates can be consistently estimated by a simple procedure based on the least-squares estimation. The proposed method is evaluated via an extensive Monte Carlo study, and the methodology is illustrated on an air pollution study.     

Understanding and classifying the role of design demonstrators in scientific exploration

This paper describes the development of a model for classifying the different type of ‘design demonstrator’ that might be used in translating scientific activity from the laboratory to the market. Two detailed case studies are described in which designers worked closely with scientists. In one of the projects, the scientists were seeking to commercialise their research. In the other, the research was at an early stage and the scientists had not considered commercialisation. Different types of physical artefact produced in these collaborative projects were analysed to identify the extent to which they might contribute to science, technology, application or market. Evidence indicates that demonstrators might fulfil multiple purposes and that the translation from science to market is more complicated than is often shown in linear models. An original classification of the role of demonstrators through this journey is provided.     

Modelling response error in school effectiveness research

Statistical modelling of school effectiveness in educational research is considered. Variance component models are generally accepted for the analysis of such studies. A shortcoming is that outcome variables are still treated as measured without an error. Unreliable variables produce biases in the estimates of the other model parameters. The variability of the relationships across schools and the effects of schools on students' outcomes differ substantially when taking the measurement error in the dependent variables of the variance component models into account. The random effects model can be extended to handle measurement error using a response model, leading to a random effects item response theory model. This extended random effects model is in particular suitable when subjects are measured repeatedly on the same outcome at several points in time.     

To each his own: Matching different entrepreneurial models to the academic scientist's individual needs

This paper presents a comparative case study of academic group leaders, active in three different scientific fields at a leading Swiss technical university. It examines the obstacles that prevent scientists from commercializing their technologies and how they can be reduced. Traditional models of technology transfer assume that scientists prefer either to 'go it alone' and become entrepreneurs (the inventor entrepreneur model) or to let go of their technologies to people interested in their commercialization (the surrogate entrepreneur model). The results of qualitative research suggest that these two models capture the extremes of a continuum populated by a variety of intermediate situations where scientists are unwilling completely to let go of their findings, but also do not want to become full time entrepreneurs. This results in considerable commercial potential that is unexploited. The Founding Angels approach might be a solution to this problem; it is designed for academics in these intermediate situations. The study contributes to the literature on university-industry technology transfer and should be useful for practitioners and scientists interested in maximizing the synergies between academia and industry.     

Chaos theory and social science: A methodological analysis

This article investigates the relevance of chaos theory for social science. The application of chaos models in the analysis of social phenomena is accompanied by some important scientific problems. First, whether observations of social phenomena are generated by nonlinear dynamics cannot be ascertained beyond considerable doubt, especially when these observations contain measurement errors; i.e., there is a problem of external validity. Secondly, and more important, as a theory of irregular cyclical social behaviour is lacking, inductive-statistical theoryformation about such behaviour, which is based on fitting a mathematical model of chaos to observations of social phenomena, is impossible unless additional information is used concerning the context and circumstances wherein the social phenomena occur; i.e., the internal validity of any theoretical explanation that is derived from only a fitted mathematical model (of chaos) cannot be assessed. So, research into the suggestion derived from mathematical chaos theory that irregular cycles may be present in the development of social phenomena over time requires theory-formation about irregular cyclical social behaviour on the basis of established theoretical insights and empirical evidence instead of fitting sophisticated mathematical models of chaos to observations of social phenomena.     

Information theoretic methods of spatial model building: A guide to the unbiased estimation of the form of probability distributions2

The methods of maximum entropy and minimum information are employed in order to provide unbiased a priori estimates of the form of probability distributions, given limited prior knowledge. These a priori estimates, in turn, can be cast in the form of theoretical models of the phenomena of interest; models to be tested against real world data. As such, the methods of maximum entropy and minimum information have relevance to any scientific or social-scientific discipline wherein the mathematical modelling of probability distributions is a concern. At the same time, however, the methods are mathematically difficult and conceptually demanding. This paper therefore presents a pedagogic introduction to the methods of maximum entropy and minimum information in the hope of making them more accessible to those social scientists who might find them relevant to their work. Using a simple example, the paper works through the two methods, one at a time, adding more and more information to the estimation problem as the discussion proceeds. The focus herein is a geographic one, yet the example employed is easily transferable to alternative contexts.     

Socio‐economic distance and spatial patterns in unemployment

This paper examines the spatial patterns of unemployment in Chicago between 1980 and 1990. We study unemployment clustering with respect to different social and economic distance metrics that reflect the structure of agents' social networks. Specifically, we use physical distance, travel time, and differences in ethnic and occupational distribution between locations. Our goal is to determine whether our estimates of spatial dependence are consistent with models in which agents' employment status is affected by information exchanged locally within their social networks. We present non‐parametric estimates of correlation across Census tracts as a function of each distance metric as well as pairs of metrics, both for unemployment rate itself and after conditioning on a set of tract characteristics. Our results indicate that there is a strong positive and statistically significant degree of spatial dependence in the distribution of raw unemployment rates, for all our metrics. However, once we condition on a set of covariates, most of the spatial autocorrelation is eliminated, with the exception of physical and occupational distance. Racial and ethnic composition variables are the single most important factor in explaining the observed correlation patterns. Copyright © 2002 John Wiley & Sons, Ltd.     

STATISTICS IN THE SOCIAL AND BEHAVIORAL SCIENCES

This survey paper gives an impression of the main ways in which statistics is used in disciplines such as sociology and psychology. After an introductory section 1 the negative image of social science research is discussed in section 2. Section 3 is devoted to the enormous influence of modern computing facilities. Measurement of human behaviour has its specific problems (section 4). The use of linear and log–linear models is the topic of section 5. Latent variables are a basic concept for social and behavioral scientists, both in some linear models (section 6) and in item response theory (section 7). In the next section multidimensional and optimal scaling techniques are mentioned, and a selection of other topics is the content of section 9. Some general remarks on the generalizability claim of statistical methods constitute the final section. Because of space limitations and priority considerations, the author has decided to write a paper about topics and not about individual research contributions. For this reason there is no list of references (it would take several pages) and no painful split of all Dutch authors into those mentioned and those omitted. In general the Dutch research community has made quite a few major contributions to the area discussed in this paper.