Risk and return measures for models of funds are typically statistics such as mean and standard deviation. A straightforward way to get a commensurable set of measures is to choose only statistics measured in the same units as the data. Often the literature uses statistics such as variance, semivariance, skewness and kurtosis not in the same units (Wilkens and Zhu, 2001; Gregoriou et al., 2005b; Nguyen-Thi-Thanh, 2006). There are several plausible explanations. First, funds returns are usually expressed as proportions and so it may not be immediately obvious that the efficiencies are not dimensionless. Second, dea models are often expressed in dual form (like model (5)–(9)) in.
