Quantile distribution approach to the one-sample location problem

Most statistical procedures regarding the mean or median may only give valid results when particular conditions about the shape of the data are met. An alternative test is presented in this thesis that does not require certain data shapes but can perform as well or better than the existing tests.

Many-sample location and scale tests with quantile-function error distributions /by Azmeri Khan

This thesis, using a computer simulation, studies the effect of the normal distribution assumption on the power of several many-sample location and scale test procedures. It also suggests an almost robust parametric test, namely numerical likelihood ratio test (NLRT) for non-normal situations. The NLRT is found better than all of the tests considered. Some real life data sets were used as examples.

OWA operators in linear regression and detection of outliers

We consider the use of Ordered Weighted Averaging (OWA) in linear regression. Our goal is to replace the traditional least squares, least absolute deviation, and maximum likelihood criteria with an OWA function of the residuals. We obtain several high breakdown robust regression methods as special cases (least median, least trimmed squares, trimmed likelihood methods). We also present new formulations of regression problem. OWA-based regression is particularly useful in the presence of outliers.

OWA operators in regression problems

We consider an application of fuzzy logic connectives to statistical regression. We replace the standard least squares, least absolute deviation, and maximum likelihood criteria with an ordered weighted averaging (OWA) function of the residuals. Depending on the choice of the weights, we obtain the standard regression problems, high-breakdown robust methods (least median, least trimmed squares, and trimmed likelihood methods), as well as new formulations. We present various approaches to numerical solution of such regression problems. OWA-based regression is particularly useful in the presence of outliers, and we illustrate the performance of the new methods on several instances of linear regression problems with multiple outliers.

Sequential fixed-width confidence bands for kernel regression estimation

We consider a random design model based on independent and identically distributed (iid) pairs of observations (Xi, Yi), where the regression function m(x) is given by m(x) = E(Yi|Xi = x) with one independent variable. In a nonparametric setting the aim is to produce a reasonable approximation to the unknown function m(x) when we have no precise information about the form of the true density, f(x) of X. We describe an estimation procedure of non-parametric regression model at a given point by some appropriately constructed fixed-width (2d) confidence interval with the confidence coefficient of at least 1−. Here, d(> 0) and 2 (0, 1) are two preassigned values. Fixed-width confidence intervals are developed using both Nadaraya-Watson and local linear kernel estimators of nonparametric regression with data-driven bandwidths.

The sample size was optimized using the purely and two-stage sequential procedure together with asymptotic properties of the Nadaraya-Watson and local linear estimators. A large scale simulation study was performed to compare their coverage accuracy. The numerical results indicate that the confidence bands based on the local linear estimator have the best performance than those constructed by using Nadaraya-Watson estimator. However both estimators are shown to have asymptotically correct coverage properties.

Two-stage sequential procedure for nonparametric regression estimation

In nonparametric statistics the functional form of the relationship between the response variable and its associated predictor variables is unspecified but it is assumed to be a smooth function. We develop a procedure for constructing a fixed width confidence interval for the predicted value at a specified point of the independent variable. The optimal sample size for constructing this interval is obtained using a two stage sequential procedure which relies on some asymptotic properties of the Nadaraya--Watson and local linear estimators. Finally, a large scale simulation study demonstrates the applicability of the developed procedure for small and moderate sample sizes. The procedure developed here should find wide applicability since many practical problems which arise in industry involve estimating an unknown function.

Sample size determination for kernel regression estimation using sequential fixed-width confidence bands

We consider a random design model based on independent and identically distributed pairs of observations (Xi, Yi), where the regression function m(x) is given by m(x) = E(Yi|Xi = x) with one independent variable. In a nonparametric setting the aim is to produce a reasonable approximation to the unknown function m(x) when we have no precise information about the form of the true density, f(x) of X. We describe an estimation procedure of non-parametric regression model at a given point by some appropriately constructed fixed-width (2d) confidence interval with the confidence coefficient of at least 1−. Here, d(> 0) and 2 (0, 1) are two preassigned values. Fixed-width confidence intervals are developed using both Nadaraya-Watson and local linear kernel estimators of nonparametric regression with data-driven bandwidths. The sample size was optimized using the purely and two-stage sequential procedures together with asymptotic properties of the Nadaraya-Watson and local linear estimators. A large scale simulation study was performed to compare their coverage accuracy. The numerical results indicate that the confi dence bands based on the local linear estimator have the better performance than those constructed by using Nadaraya-Watson estimator. However both estimators are shown to have asymptotically correct coverage properties.

Modelling for injection moulding process based on principal component regression method

Determination of the optimal operating condition for moulding process has been of special interest for many researchers. To determine the optimal setting, one has to derive the model of injection moulding process first which is able to map the relationship between the input process control factors and output responses. One of most popular modeling techniques is the linear least square regression due to its effectiveness and completeness. However, the least square regression was found to be very sensitive to the outliers and failed to provide a reliable model if the control variables are highly related with each other. To address this problem, a new modeling method based on principal component regression was proposed in this paper. The distinguished feature of our proposed method is it does not only consider the variance of covariance matrix of control variables but also consider the correlation coefficient between control variables and target variables to be optimised. Such a modelling method has been implemented into a commercial optimisation software and field test results demonstrated the performance of the proposed modelling method.

Assessment of spatiotemporal varying relationships between rainfall, land cover and surface water area using geographically weighted regression

Traditional regression techniques such as ordinary least squares (OLS) are often unable to accurately model spatially varying data and may ignore or hide local variations in model coefficients. A relatively new technique, geographically weighted regression (GWR) has been shown to greatly improve model performance compared to OLS in terms of higher R 2 and lower corrected Akaike information criterion (AICC). GWR models have the potential to improve reliabilities of the identified relationships by reducing spatial autocorrelations and by accounting for local variations and spatial non-stationarity between dependent and independent variables. In this study, GWR was used to examine the relationship between land cover, rainfall and surface water habitat in 149 sub-catchments in a predominately agricultural region covering 2.6 million ha in southeast Australia. The application of the GWR models revealed that the relationships between land cover, rainfall and surface water habitat display significant spatial non-stationarity. GWR showed improvements over analogous OLS models in terms of higher R 2 and lower AICC. The increased explanatory power of GWR was confirmed by the results of an approximate likelihood ratio test, which showed statistically significant improvements over analogous OLS models. The models suggest that the amount of surface water area in the landscape is related to anthropogenic drainage practices enhancing runoff to facilitate intensive agriculture and increased plantation forestry. However, with some key variables not present in our analysis, the strength of this relationship could not be qualified. GWR techniques have the potential to serve as a useful tool for environmental research and management across a broad range of scales for the investigation of spatially varying relationships.

A panel regression method for quantifying the convergence of house prices

The convergence among house prices has attracted much attention from researchers. Previous research mainly utilised a time-series regression method to investigate convergences of house prices, which may ignore the heterogeneity of houses across cities. This research developed a panel regression method, by which the heterogeneity of house prices can be captured. Seemingly unrelated regression estimators were also adapted to deal with the contemporary correlations across cities. Investigation of the convergence among house prices in the Australian capital cities was carried out by using the developed panel regression method. Results suggested that house prices converge in Sydney, Adelaide and Hobart but diverge in Darwin.