You are here:
Publication details
Kernel Regression Model for Total Ozone Data
Authors | |
---|---|
Year of publication | 2013 |
Type | Article in Periodical |
Magazine / Source | Journal of Environmental Statistics |
MU Faculty or unit | |
Citation | |
Field | General mathematics |
Keywords | total ozone; kernel; bandwidth selection |
Description | The present paper is focused on an analysis of total ozone data from Dobson spectrophotometer operating at the Vernadsky station, the British Antarctic Survey (BAS), former Faraday station. The data sets were processed as time points measuring amount of ozone. In order to analyze such data we propose a kernel regression method. Kernel methods represent one of the most effective nonparametric methods. But there is a serious difficulty connected with them -- the choice of a smoothing parameter called a bandwidth. In the case of independent observations the literature on bandwidth selection methods is quite extensive. However, if the observations are dependent, then classical bandwidth selectors have not always provided applicable results. There exist several possibilities for overcoming the effect of dependence on bandwidth selection. In the present paper we use the results of Chu, Marron (1991) and Kolacek (2008) and develop two methods for the bandwidth choice. We apply the above mentioned methods to the time series of ozone data. All discussed methods are implemented in Matlab. |