A robust estimate for location in experiments with plantation crops

Abstract

Presence of outliers or extreme values in the experimental data is a major concern for data analysis. Many times, the experimental data contains abnormal or extreme values due to genetic variations, loss of yield due to pest/disease infestation, errors in tabulation/data entry etc. In field experimental data analysis, the plot mean or treatment mean is usually considered for comparison. The sample mean, which is usually taken as an estimate of the population mean (location estimator) is highly affected by the presence of outliers or extreme values particularly when the sample size is small. In this paper, kernel weighted location estimator with weight proportional to the value of the estimated kernel density function is proposed, to handle the outliers or extreme values. The kernel weighted location estimator is robust even if the underlying distribution is non-symmetric. The performance of the proposed method is compared with the existing procedures through simulation study. The method is also applied to the analysis of yield data of arecanut and cocoa mixed cropping experiment at CPCRI Regional Station, Vittal.