An assurance interval for the non-Archimedean epsilon in data envelopment analysis (DEA) models

This paper clarifies the role of non-Archimedean infinitesimal ε in DEA models so that the associated linear programs may be infeasible (for the multiplier side) and unbounded (for the envelopment side) for certain values of ε. It is shown that the bound of ε proposed by Ali and Seiford is invalid for feasibility and boundedness of the linear programs. A procedure is presented for determining an assurance interval of ε. It is also shown that an assurance value for ε can be found using a single linear program.