Abstract
dc:description.abstractEnvironmental Impact Assessment consists of assessing the possible effects of a project of development on the environment. Statistical methods which are used in this process include the ability to detect the magnitude of an impact at the time in which an intervention has occurred, and the ability to detect when an intervention has occurred if the time of impact is not available. This impact at the time at which the intervention has occurred is also known as a ‘changepoint’. This project consists of evaluating methods used to assess the magnitude of these ‘impacts’, and the methods which can be used to determine where an impact has occurred if the time of impact is not available. A number of datasets have been made available for analysis within this thesis, each of these datasets are expected to contain impacts within them. By reviewing the techniques and methods which are available for statisticians, applications to these datasets were made. The classic approach to quantifying the magnitude of change within a series when the time of the impact is known is Before-After-Control-Impact design, which uses a dummy variable within a linear regression model to determine whether a significant change in mean can be detected. This thesis reviewed the history and the approaches used by statisticians when using this method, and the adaptions that can be made. BACI was then applied to our own datasets to determine whether a significant change in mean could be detected. Many statistical changepoint detection methods are available to statisticians when the location of a changepoint, if any, is not known. By summarising the methods available and calculating the power of various methods via a simulation study, a number of changepoint detection methods were applied to real life data. Finally, various modelling techniques were applied to the available datasets and by incorporating terms to indicate the detected location of changepoints, we could determine whether adding these terms gives better fitting models.
Degree
thesis:*- Level dc:type.qualificationlevel
- MSc(R)
- Grantor dc:publisher.institution
- University of Glasgow
- Year dc:date.issued
- 2013
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Tough, Fraser
Subjects
dc:subject × 3Rights
- Language dc:language
- en