A diagnostic tool for detecting outliers in experimental data

Article type: Research Article

Authors: Bhar, Lalmohan

Affiliations: Indian Agricultural Statistics Research Institute, Pusa, New Delhi, India. E-mail: lmbhar@gmail.com; lmbhar@iasri.res.in

Abstract: Cook-statistic has been developed for identifying outliers in experimental data. A general formula has been obtained for detection of any k outliers. A particular case of a single outlier has also been considered. The derived statistic has been applied to real experimental data. Upon deletion of the identified outlying observation, it has been observed that the conclusion to be drawn from the experiment is reversed.

Keywords: Block design, Cook-statistic, mustard, outlier

DOI: 10.3233/MAS-2012-0237

Journal: Model Assisted Statistics and Applications, vol. 8, no. 1, pp. 61-68, 2013