Title: Detecting outliers in DEA after correcting biases in efficiency scores using simulated beta samples

Authors: Parakramaweera Sunil Dharmapala

Addresses: Department of Mathematics, Lone Star College, Cy-Fair Campus, Cypress, Texas 77433, USA

Abstract: We address the issue that data used in DEA are possibly contaminated with statistical noise, and as a result, efficiency scores deviate from actual values due to statistical bias. In such situation, detecting outliers could be misleading. Therefore, we propose a method to correct the biases and compute DEA efficiency scores that follow underlying beta distributions, and we use Thompson et al.'s (1996) DEA model with assurance regions. Then, we demonstrate the bias-correction process of the DEA frontier estimates by constructing confidence intervals for the mean scores and use them to detect outliers.

Keywords: data envelopment analysis; DEA; assurance regions; order statistics; beta distribution; bias-correction; outliers.

DOI: 10.1504/IJMOR.2021.116961

International Journal of Mathematics in Operational Research, 2021 Vol.19 No.3, pp.387 - 400

Received: 01 Feb 2020
Accepted: 15 May 2020
Published online: 10 Aug 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article