CMStatistics 2018: Start Registration
View Submission - CFE
Title: Nonparametric trend Universal kriging method with applications to air quality data Authors:  Lifeng Yang - University of Southampton (United Kingdom) [presenting]
Zudi Lu - University of Southampton (United Kingdom)
Abstract: The three commonly used linear kriging methods, i.e. simple -, ordinary- and universal kriging, offer the property of best linear unbiased predictor. However, these uncomplicated spatial prediction and mapping procedures, which make them appealing to many practitioners, are subject to criticism owing mainly to their over-simplified linear regression trend structure, often introducing misspecifications to its spatial function. Instead of linear trend structure of data we propose a Nonparametric-Trend Universal Kriging (NTUK) method, to overcome this structural drawback by delegating the estimation of spatial trend to a nonparametric function. Combining its result with the current Universal kriging method, the new NTUK acts as an improved alternative to the existing linear methods. Asymptotic justification for this estimation procedure is developed. Empirical applications of the above methods to air quality in the UK are compared to show the improvement of the proposed NTUK prediction.