Use of progression-free survival in the evaluation of clinical interventions is hampered by a variety of issues, including censoring patterns not addressed in the usual methods for survival analysis. Progression can be right-censored before survival or interval-censored between inspection times. Current practice calls for imputing events to their time of detection. Such an approach is prone to bias, underestimates standard errors and makes inefficient use of the data at hand. Moreover a composite outcome prevents inference about the actual treatment effect on the risk of progression. This thesis develops semiparametric and sieve maximum likelihood estimators to more formally analyze progression-related endpoints. For the special case where death rarely precedes progression, a Cox-Aalen model is proposed for regression analysis of time-to-progression under intermittent inspection. The general setting considering both progression and survival is examined with a Markov Cox-type illness-death model under various censoring schemes. All of the resulting estimators globally converge to the truth slower than the parametric rate, but their finite-dimensional components are asymptotically efficient. Numerical studies suggest that the new methods perform better than their imputation-based alternatives under moderate to large samples having higher rates of censoring.

The thesis consists of four papers that address likelihood-based unit root tests for panel data with cross-sectional dependence arising from common factors. In the first three papers, we derive Lagrange multiplier (LM)-type tests for common and idiosyncratic unit roots in the exact factor models based on the likelihood function of the differenced data. Also derived are the asymptotic distributions of these test statistics. The finite sample properties of these tests are compared by simulation with other commonly used unit root tests. The results show that our LM-type tests have better size and local power properties. In the fourth paper, we estimate the spaces spanned by the common factors and the spaces spanned by the idiosyncratic components of the static factor model by using the quasi-maximum likelihood (ML) method and compare it with the widely used method of principal components (PC). Next, by simulation, we compare the size and power properties of established tests for idiosyncratic unit roots, using both the ML and PC methods. Simulation results show that the idiosyncratic unit root tests based on the likelihood-based residuals generally have better size and higher size-adjusted power, especially when the cross-sectional dimension is small and the time series dimension is large.

One-way and two-way classification models in experimental design for both balanced and unbalanced cases are considered when the errors have Generalized Secant Hyperbolic distribution. Efficient and robust estimators for main and interaction effects are obtained by using the modified maximum likelihood estimation (MML) technique. The test statistics analogous to the normal-theory F statistics are defined to test main and interaction effects and a test statistic for testing linear contrasts is defined. It is shown that test statistics based on MML estimators are efficient and robust. The methodogy obtained is also generalized to situations where the error distributions from block to block are non-identical.

In the first part of this study, two entropy methods under different distribution assumptions are examined on a network of stream gauging stations located in Kizilirmak Basin to rank the stations according to their level of importance. The stations are ranked by using two different entropy methods under different distributions. Thus, showing the effect of the distribution type on both entropy methods is aimed. In the second part of this study, autoregressive models with asymmetric innovations and an artificial neural network model are introduced. Autoregressive models (AR) which have been developed in hydrology are based on several assumptions. The normality assumption for the innovations of AR models is investigated in this study. The main reason of making this assumption in the autoregressive models established is the difficulties faced in finding the model parameters under the distributions other than the normal distributions. From this point of view, introduction of the modified maximum likelihood procedure developed by Tiku et. al. (1996) in estimation of the autoregressive model parameters having non-normally distributed residual series, in the area of hydrology has been aimed. It is also important to consider how the autoregressive model parameters having skewed distributions could be estimated. Besides these autoregressive models, the artificial neural network (ANN) model was also constructed for annual and monthly hydrologic time series due to its advantages such as no statistical distribution and no linearity assumptions. The models considered are applied to annual and monthly streamflow data obtained from five streamflow gauging stations in Kizilirmak Basin. It is shown that AR(1) model with Weibull innovations provides best solutions for annual series and AR(1) model with generalized logistic innovations provides best solution for monthly as compared with the results of artificial neural network models.

国立情報学研究所で電子化したコンテンツを使用している。

Stochastic di®erential equations (SDEs) are　central to much of modern finance theory and have been widely used to model the behaviour of key variables such as the　instantaneous short-term interest rate, asset prices, asset returns and their volatility.　The explanatory and/or predictive power of these models depends crucially on the particularisation of the model SDE(s) to real data through the choice of values for their parameters. In econometrics, optimal parameter estimates are generally considered to　be those that maximise the likelihood of the sample. In the context of the estimation of the parameters of SDEs, however, a closed-form expression for the likelihood　function is rarely available and hence exact maximum-likelihood (EML) estimation is usually infeasible. The key research problem examined in this thesis is the development of generic, accurate and computationally feasible estimation procedures based on the　ML principle, that can be implemented in the absence of a closed-form expression for　the likelihood function. The overall recommendation to come out of the thesis is that　an estimation procedure based on the finite-element solution of a reformulation of the Fokker-Planck equation in terms of the transitional cumulative distribution function(CDF) provides the best balance across all of the desired characteristics. The recommended approach involves the use of an interpolation technique proposed in this thesis which greatly reduces the required computational effort.

As well as the many benefits associated with the evolution of multispectral sensors into hyperspectral sensors there is also a considerable increase in storage space and the computational load to process the data. Consequently the remote sensing ommunity is investigating and developing statistical methods to alleviate these problems. / The research presented here investigates several approaches to reducing the dimensionality of hyperspectral remotely sensed data while maintaining the levels of accuracy achieved using the full dimensionality of the data. It was conducted with an emphasis on applications in maximum likelihood classification (MLC) of hyperspectral image data. An inherent characteristic of hyperspectral data is that adjacent bands are typically highly correlated and this results in a high level of redundancy in the data. The high correlations between adjacent bands can be exploited to realise significant reductions in the dimensionality of the data, for a negligible reduction in classification accuracy. / The high correlations between neighbouring bands is related to their response functions overlapping with each other by a large amount. The spectral band filter functions were modelled for the HyMap instrument that acquires hyperspectral data used in this study. The results were compared with measured filter function data from a similar, more recent HyMap instrument. The results indicated that on average HyMap spectral band filter functions exhibit overlaps with their neighbouring bands of approximately 60%. This is considerable and partly accounts for the high correlation between neighbouring spectral bands on hyperspectral instruments. / A hyperspectral HyMap image acquired over an agricultural region in the south west of Western Australia has been used for this research. The image is composed of 512 × 512 pixels, with each pixel having a spatial resolution of 3.5 m. The data was initially reduced from 128 spectral bands to 82 spectral bands by removing the highly overlapping spectral bands, those which exhibit high levels of noise and those bands located at strong atmospheric absorption wavelengths. The image was examined and found to contain 15 distinct spectral classes. Training data was selected for each of these classes and class spectral mean and covariance matrices were generated. / The discriminant function for MLC makes use of not only the measured pixel spectra but also the sample class covariance matrices. This thesis first examines reducing the parameterization of these covariance matrices for use by the MLC algorithm. The full dimensional spectra are still used for the classification but the number of parameters needed to describe the covariance information is significantly reduced. When a threshold of 0.04 was used in conjunction with the partial correlation matrices to identify low values in the inverse covariance matrices, the resulting classification accuracy was 96.42%. This was achieved using only 68% of the elements in the original covariance matrices. / Both wavelet techniques and cubic splines were investigated as a means of representing the measured pixel spectra with considerably fewer bands. Of the different mother wavelets used, it was found that the Daubechies-4 wavelet performed slightly better than the Haar and Daubechies-6 wavelets at generating accurate spectra with the least number of parameters. The wavelet techniques investigated produced more accurately modelled spectra compared with cubic splines with various knot selection approaches. A backward stepwise knot selection technique was identified to be more effective at approximating the spectra than using regularly spaced knots. A forward stepwise selection technique was investigated but was determined to be unsuited to this process. / All approaches were adapted to process an entire hyperspectral image and the subsequent images were classified using MLC. Wavelet approximation coefficients gave slightly better classification results than wavelet detail coefficients and the Haar wavelet proved to be a more superior wavelet for classification purposes. With 6 approximation coefficients, the Haar wavelet could be used to classify the data with an accuracy of 95.6%. For 11 approximation coefficients this figure increased to 96.1%. / First and second derivative spectra were also used in the classification of the image. The first and second derivatives were determined for each of the class spectral means and for each band the standard deviations were calculated of both the first and second derivatives. Bands were then ranked in order of decreasing standard deviation. Bands showing the highest standard deviations were identified and the derivatives were generated for the entire image at these wavelengths. The resulting first and second derivative images were then classified using MLC. Using 25 spectral bands classification accuracies of approximately 96% and 95% were achieved using the first and second derivative images respectively. These results are comparable with those from using wavelets although wavelets produced higher classification accuracies when fewer coefficients were used.

Abstract This thesis considers some linear and nonlinear time series models. In the linear case, the analysis of a large number of short time series generated by a first order autoregressive type model is considered. The conditional and exact maximum likelihood procedures are developed to estimate parameters. Simulation results are presented and compare the bias and the mean square errors of the parameter estimates. In Chapter 3, five important nonlinear models are considered and their time series properties are discussed. The estimating function approach for nonlinear models is developed in detail in Chapter 4 and examples are added to illustrate the theory. A simulation study is carried out to examine the finite sample behavior of these proposed estimates based on the estimating functions.

octor of Philosophy(PhD) / Molecular Evolution is the key to explain the divergence of species and the origin of life on earth. The main task in the study of molecular evolution is the reconstruction of evolutionary trees from sequences data of the current species. This thesis introduces a novel algorithm for inferring evolutionary trees from genetic data using quartet-based approach. The new method recursively merges sub-trees based on a global statistical provided by the global quartet weight matrix. The quarte weights can be computed using several methods. Since the quartet weights computation is the most expensive procedure in this approach, the new method enables the parallel inference of large evolutionary trees. Several techniques developed to deal with quartets inaccuracies. In addition, the new method we developed is flexible in such a way that can combine morphological and molecular phylogenetic analyses to yield more accurate trees. Also, we introduce the concept of critical point where more than one possible merges are possible for the same sub-tree. The critical point concept can provide information about the relationships between species in more details and show how close they are. This enables us to detect other reasonable trees. We evaluated the algorithm on both synthetic and real data sets. Experimental results showed that the new method achieved significantly better accuracy in comparison with existing methods.

Zugl.: Regensburg, Universiẗat, Diss., 1998.