Nonparametric methodologies for regression models with correlated data

Giannitrapani, Marco (2006) Nonparametric methodologies for regression models with correlated data. PhD thesis, University of Glasgow.

Full text available as:
[thumbnail of 10753992.pdf] PDF
Download (11MB)


In many spatial and temporal data sets, nonparametric techniques have recently been widely used because of their ability to model without requiring any assumptions on the distributional form of the data. However many nonparametric tools assume independent errors, that is not always the case. The present work extends some of the well established nonparametric techniques in order to make them applicable even with correlated data. Simulation studies will show the performances of the proposed methodologies. The methods are applied to air pollution data monitored over Europe in last quarter of the twentieth century by EMEP (Co-operative Programme for Monitoring and Evaluation of the long Range Transmission of Air Pollutants in Europe) and by OECD (Organization for Economic Co-operation and Development). Chapter 1 gives a background to the air pollution problems, introduces the questions of interest and the aims of this work. It also shows some characteristics of the data that will be necessary to take into account for the analysis that will be done in the following chapters. Chapter 2 reviews some of the existing nonparametric methodologies that, however relying on the assumption of independent errors, could be applied to the data. Chapter 3 presents a diagnostic to detect discontinuities in a one-dimensional nonparametric regression accounting for correlated errors. A simulation study shows the performance of the proposed test, and the results of its application to air pollution data (SO[2], SO[4] in air and SO[4] in precipitation) monitored across 130 sites in Europe from 1970's to 2000, will be presented. Chapter 4 presents the generalization of well established nonparametric techniques that can model and test correlated data. Simulation studies show the performances of the proposed modeling tools. Chapter 5 shows applications of the methodologies presented in Chapter 4 to air pollution data. Chapter 6 develops binned versions of the methodologies introduced in Chapter 4 allowing to fit and test models with large data sets, such as spatiotemporal ones, that show correlation. Chapter 7 presents an analysis of the relationship between the SO[2] emissions and the monitored SO[2] concentrations. Chapter 8 will summarize the main conclusions with a final discussion on possible future work.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Additional Information: Adviser: Adrian Bowman
Keywords: Statistics
Date of Award: 2006
Depositing User: Enlighten Team
Unique ID: glathesis:2006-74082
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 23 Sep 2019 15:33
Last Modified: 23 Sep 2019 15:33

Actions (login required)

View Item View Item


Downloads per month over past year