In this thesis, we try to provide a broad econometric analysis of a class of risk measures, distortion risk measures (DRM). With carefully selected functional form, the Value-at-Risk (VaR) and Tail-VaR (TVaR) are special cases of DRMs. Besides, the DRM also admits interpretation in the sense of non-expected utility type of preferences. We first provide a unified statistical framework for the nonparametric estimators of the DRMs in a univariate case. The asymptotic properties of both the DRMs and their sensitivities with respect to the parameters representing risk aversion and/or pessimism are derived. Moreover, the relationships between the VaR and TVaR are also investigated in detail, which, we hope, can shed new lights on the way passing one risk measure to another. Then, the analysis of DRMs are extended to a multi-dimensional framework, where the DRM is computed for a portfolio consisting of many primitive assets. Analogous to the mean-variance frontier analysis, we study the efficient portfolio frontier when both objective and constraint are replaced by the DRMs. We call this the DRM-DRM framework. Under a nonparametric setting, we propose three asymptotic test statistics for evaluating the efficiency of a given portfolio. Finally, we discuss the criteria used for evaluating models used to forecast the VaRs. More precisely, we propose a criterion which takes into account the loss levels beyond the VaRs.

In this thesis, we try to provide a broad econometric analysis of a class of risk measures, distortion risk measures (DRM). With carefully selected functional form, the Value-at-Risk (VaR) and Tail-VaR (TVaR) are special cases of DRMs. Besides, the DRM also admits interpretation in the sense of non-expected utility type of preferences. We first provide a unified statistical framework for the nonparametric estimators of the DRMs in a univariate case. The asymptotic properties of both the DRMs and their sensitivities with respect to the parameters representing risk aversion and/or pessimism are derived. Moreover, the relationships between the VaR and TVaR are also investigated in detail, which, we hope, can shed new lights on the way passing one risk measure to another. Then, the analysis of DRMs are extended to a multi-dimensional framework, where the DRM is computed for a portfolio consisting of many primitive assets. Analogous to the mean-variance frontier analysis, we study the efficient portfolio frontier when both objective and constraint are replaced by the DRMs. We call this the DRM-DRM framework. Under a nonparametric setting, we propose three asymptotic test statistics for evaluating the efficiency of a given portfolio. Finally, we discuss the criteria used for evaluating models used to forecast the VaRs. More precisely, we propose a criterion which takes into account the loss levels beyond the VaRs.

In this dissertation, we utilize the discrete Fourier analysis on axially symmetric data generation and nonparametric estimation. We first represent the axially symmetric process as Fourier series on circles with the Fourier random coefficients expressed as circularlysymmetric complex random vectors. We develop an algorithm to generate the axially symmetric data that follow the given covariance function. Our simulation study demonstrates that our approach performs comparable with the classical approach using the given axially symmetric covariance function directly, while at the same time significantly reducing computational costs. For the second contribution of this dissertation, we apply the discrete Fourier transform to provide the nonparametric estimation on the covariance function of the above circularly-symmetric complex random vectors under gridded data structure. Our results show that these estimates has closely related to the simultaneous diagonalization of circulant matrices. The simulation study shows that our proposed estimates match well with their theoretical counterparts. Finally through the Fourier transform of the original gridded data, the covariance estimator of an axially symmetric process based on the method of moments can be represented as a quadratic form of transformed data that is associated with a rotation matrix.

京都大学 / 0048 / 新制・課程博士 / 博士(経済学) / 甲第19164号 / 経博第518号 / 新制||経||274(附属図書館) / 32156 / 京都大学大学院経済学研究科経済学専攻 / (主査)教授 西山 慶彦, 准教授 奥井 亮, 准教授 山田 憲 / 学位規則第4条第1項該当 / Doctor of Economics / Kyoto University / DFAM

The nonparametric confidence interval for an unknown function is quite a useful tool in statistical inferential procedures; and thus, there exists a wide body of literature on the topic. The primary issues are the smoothing parameter selection using an appropriate criterion and then the coverage probability and length of the associated confidence interval. Here our focus is on the interval length in general and, in particular, on the variability in the lengths of nonparametric intervals for probability density and hazard rate functions. We start with the analysis of a nonparametric confidence interval for a probability density function noting that the confidence interval length is directly proportional to the square root of a density function. That is variability of the length of the confidence interval is driven by the variance of the estimator used to estimate the square-root of the density function. Therefore we propose and use a kernel-based constant variance estimator of the square-root of a density function. The performance of confidence intervals so obtained is studied through simulations. The methodology is then extended to nonparametric confidence intervals for the hazard rate function. Changing direction somewhat, the second part of this thesis presents a statistical study of daily snow trends in the United States and Canada from 1960-2009. A storage model balance equation with periodic features is used to describe the daily snow depth process. Changepoint (inhomogeneities features) are permitted in the model in the form of mean level shifts. The results show that snow depths are mostly declining in the United States. In contrast, snow depths seem to be increasing in Canada, especially in north-western areas of the country. On the whole, more grids are estimated to have an increasing snow trend than a decreasing trend. The changepoint component in the model serves to lessen the overall magnitude of the trends in most locations.

Cópulas tem se tornado uma importante ferramenta para descrever e analisar a estrutura de dependência entre variáveis aleatórias e processos estocásticos. Recentemente, surgiram alguns métodos de estimação não paramétricos, utilizando kernels e ondaletas. Neste contexto, sabendo que cópulas podem ser escritas como expansão em ondaletas, foi proposto um estimador não paramétrico via ondaletas para a função cópula para dados independentes e de séries temporais, considerando processos alfa-mixing. Este estimador tem como característica principal estimar diretamente a função cópula, sem fazer suposição alguma sobre a distribuição dos dados e sem ajustes prévios de modelos ARMA - GARCH, como é feito em ajuste paramétrico para cópulas. Foram calculadas taxas de convergência para o estimador proposto em ambos os casos, mostrando sua consistência. Foram feitos também alguns estudos de simulação, além de aplicações a dados reais. / Copulas are important tools for describing the dependence structure between random variables and stochastic processes. Recently some nonparametric estimation procedures have appeared, using kernels and wavelets. In this context, knowing that a copula function can be expanded in a wavelet basis, we have proposed a nonparametric copula estimation procedure through wavelets for independent data and times series under alpha-mixing condition. The main feature of this estimator is the copula function estimation without assumptions about the data distribution and without ARMA - GARCH modeling, like in parametric copula estimation. Convergence rates for the estimator were computed, showing the estimator consistency. Some simulation studies were made, as well as analysis of real data sets.

Diferente das medidas de associação global (coeficiente de correlação linear de Pearson, de Spearman, tau de Kendall, por exemplo), as medidas de dependência local descrevem o comportamento da dependência localmente em diferentes regiões. Nesta tese, as medidas de dependência local para variáveis aleatórias propostas por Bairamov et al. (2003), Bjerve e Doksum (1993) e Sibuya (1960), são estudadas sob o enfoque de processos estocásticos estacionários bivariados e univariados, neste caso, estudando o comportamento da dependência local ao longo das defasagens da série temporal. Para as duas primeiras medidas, discutimos as suas propriedades, e estudamos os seus estimadores, além da consistência dos mesmos. Para a medida de Sibuya, além de discutir suas propriedades, propomos três estimadores para variáveis aleatórias e dois para séries temporais, verificando a consistência dos mesmos. O comportamento das três medidas locais e dos seus estimadores foram avaliados através de simulações e aplicações a dados reais (neste caso, fizemos uma comparação destas com cópula e densidade cópula). / Unlike global association measures (Pearson´s linear correlation coefficient, Spearman´s rho, Kendall´s tau, for example), local dependence measures describe the behaviour of dependence locally in different regions. In this thesis, the local dependence measures for random variables proposed by Bairamov et al. (2003), Bjerve and Doksum (1993) and Sibuya (1960), are studied in the context of bivariate and univariate stationary stochastic processes, in this case, evaluating the performance of local dependence along time lags. We discussed the properties and studied the estimators and consistence of the first two measures. As for the Sibuya measure, in addition to discussing its properties, we propose three estimators for random variables and two for time series while checking their consistence. The behaviour of the three local measures and their respective estimators was evaluated by simulations and application to real data (in this case, a comparison was drawn with copula and copula density).

In this dissertation, we extend several relatively new developments in statistical model selection and data mining in order to improve one of the workhorse statistical tools - mixture modeling (Pearson, 1894). The traditional mixture model assumes data comes from several populations of Gaussian distributions. Thus, what remains is to determine how many distributions, their population parameters, and the mixing proportions. However, real data often do not fit the restrictions of normality very well. It is likely that data from a single population exhibiting either asymmetrical or nonnormal tail behavior could be erroneously modeled as two populations, resulting in suboptimal decisions. To avoid these pitfalls, we develop the mixture model under a broader distributional assumption by fitting a group of multivariate elliptically-contoured distributions (Anderson and Fang, 1990; Fang et al., 1990). Special cases include the multivariate Gaussian and power exponential distributions, as well as the multivariate generalization of the Student’s T. This gives us the flexibility to model nonnormal tail and peak behavior, though the symmetry restriction still exists. The literature has many examples of research generalizing the Gaussian mixture model to other distributions (Farrell and Mersereau, 2004; Hasselblad, 1966; John, 1970a), but our effort is more general. Further, we generalize the mixture model to be non-parametric, by developing two types of kernel mixture model. First, we generalize the mixture model to use the truly multivariate kernel density estimators (Wand and Jones, 1995). Additionally, we develop the power exponential product kernel mixture model, which allows the density to adjust to the shape of each dimension independently. Because kernel density estimators enforce no functional form, both of these methods can adapt to nonnormal asymmetric, kurtotic, and tail characteristics. Over the past two decades or so, evolutionary algorithms have grown in popularity, as they have provided encouraging results in a variety of optimization problems. Several authors have applied the genetic algorithm - a subset of evolutionary algorithms - to mixture modeling, including Bhuyan et al. (1991), Krishna and Murty (1999), and Wicker (2006). These procedures have the benefit that they bypass computational issues that plague the traditional methods. We extend these initialization and optimization methods by combining them with our updated mixture models. Additionally, we “borrow” results from robust estimation theory (Ledoit and Wolf, 2003; Shurygin, 1983; Thomaz, 2004) in order to data-adaptively regularize population covariance matrices. Numerical instability of the covariance matrix can be a significant problem for mixture modeling, since estimation is typically done on a relatively small subset of the observations. We likewise extend various information criteria (Akaike, 1973; Bozdogan, 1994b; Schwarz, 1978) to the elliptically-contoured and kernel mixture models. Information criteria guide model selection and estimation based on various approximations to the Kullback-Liebler divergence. Following Bozdogan (1994a), we use these tools to sequentially select the best mixture model, select the best subset of variables, and detect influential observations - all without making any subjective decisions. Over the course of this research, we developed a full-featured Matlab toolbox (M3) which implements all the new developments in mixture modeling presented in this dissertation. We show results on both simulated and real world datasets. Keywords: mixture modeling, nonparametric estimation, subset selection, influence detection, evidence-based medical diagnostics, unsupervised classification, robust estimation.

This dissertation is composed of four articles describing inference and visualization of periodic sequences. In the first article, a nonparametric method is proposed for estimating the period and values of a periodic sequence when the data are evenly spaced in time. The period is estimated by a "leave-out-one-cycle" version of cross-validation (CV) and complements the periodogram, a widely used tool for period estimation. The CV method is computationally simple and implicitly penalizes multiples of the smallest period, leading to a "virtually" consistent estimator. The second article is the multivariate extension, where we present a CV method of estimating the periods of multiple periodic sequences when data are observed at evenly spaced time points. The basic idea is to borrow information from other correlated sequences to improve estimation of the period of interest. We show that the asymptotic behavior of the bivariate CV is the same as the CV for one sequence, however, for finite samples, the better the periods of the other correlated sequences are estimated, the more substantial improvements can be obtained. The third article proposes an informative exploratory tool, the functional boxplot, for visualizing functional data, as well as its generalization, the enhanced functional boxplot. Based on the center outwards ordering induced by band depth for functional data, the descriptive statistics of a functional boxplot are: the envelope of the 50 percent central region, the median curve and the maximum non-outlying envelope. In addition, outliers can be detected by the 1.5 times the 50 percent central region empirical rule. The last article proposes a simulation-based method to adjust functional boxplots for correlations when visualizing functional and spatio-temporal data, as well as detecting outliers. We start by investigating the relationship between the spatiotemporal dependence and the 1.5 times the 50 percent central region empirical outlier detection rule. Then, we propose to simulate observations without outliers based on a robust estimator of the covariance function of the data. We select the constant factor in the functional boxplot to control the probability of correctly detecting no outliers. Finally, we apply the selected factor to the functional boxplot of the original data.

This dissertation contains two essays studying panel data econometric models. First, we consider the problem of estimating a nonparametric panel data models with fixed effects. We propose using the profile least squares method to concentrate out the fixed effects and then estimate the unknown function by the kernel method. We show that our proposed estimator is consistent and has an asymptotically normal distribution. Monte Carlo simulations show that our proposed estimator performs well compared with several existing estimators. Second, we study the effects of Hong Kong’s fixed exchange rate against U.S. dollar using a novel panel data method. After the 1997 Asian Financial Crisis, many of the Asia countries adopted flexible exchange rate policies while Hong Kong still keeps its fixed exchange rate. By comparing Hong Kong versus its major trading partners, we show that if, like other Asian countries, Hong Kong had adopted a float exchange rate policy in October 1998, Hong Kong’s (counterfactual) total value of exports would increase by 14.65 %. Similarly, Hong Kong’s total value of imports would increase about 31%. We conclude that Hong Kong dollar is overvalued by 9.34% due to its fixed exchange rate policy.