An application of covariance analysis to capacity expansion at an inter-modal blending plant

Analysis of covariance (ANCOVA) integrates analysis of variance (ANOVA) and regression. The basic advantages of ANCOVA over ANOVA are: (1) generally greater power, and (2) reduction in bias caused by differences between groups that exist before experimental treatments are administered. ANCOVA has numerous possible applications in the evaluation of simulation output, especially where the values of covariates are not known until after the simulation experiment is completed. These covariates are uncontrolled experimental variables that influence the response but are themselves unaffected by the experimental factors. This paper provides an application of multiple analysis of covariance (MANCOVA) to a simulation experiment to determine whether an intermodal transfer and bending facility should add commodity handling and storage capacity. A discrete simulation model of the plant generated cash flows from several proposed capital projects. These cash flows indicated that capacity expansion was a prudent decision. However, when the treatment means for the various combinations of additional capacity were adjusted by MANCOVA for the same levels of operating volume and scheduling performance, the adjusted cash flows produced unacceptable financial returns. In this example, the increased precision of the MANCOVA model suggested that plant management should not invest in additional storage and commodity handling capacity.