Subsampling bootstrap in network DEA

• Subsampling bootstrap procedure for DEA estimator is extended to network structure The subsampling proposed also considers undesirable factors Evidence on the performance of obtained through Monte Carlo experiments method applied evaluation railways in OECD considering noise pollution problem Data Envelopment Analysis (DEA), provides an empirical estimation production frontier, based observed sample decision making units (DMUs). Except single input-single output case, asymptotic distribution can only be approximated bootstrapping approaches. Therefore, techniques have been widely literature make statistical inference cases when process has a single-stage structure. However, many cases, transformation inputs into outputs inner that needs considered. This paper examines applicability approximation structure, and presence factors. case two-stage series structures, where overall stage efficiency scores are calculated using additive decomposition approach. Results indicate great sensitivity both subsample size, as well data generating process. methodology then construct confidence interval estimates 22 European countries, railway transport decomposed two stages considered output.