The potential variables that affect energy efficiency were first
entered in a stepwise regression to identify those variables that are
statistically significant predictors of EPI and AAhEPI. This analysis
was carried out using the Statistical Analysis Software (SAS v9.0).
Only those variables that were significant and had the expected
sign of the coefficient were retained. The scatter plots of the significant
operational characteristics were plotted against the dependent
variables to determine whether the relationships were linear or
logarithmic. The significant variables that represent the operational
characteristics, linear or logarithmic, were standardised and regression
coefficients were obtained for these from a multiple linear
regression with the dependent variables. This method has been
adopted by Chung et al. [17] to benchmark the Energy Use Intensity
(EUI) of 30 randomly selected supermarkets in Hong Kong. Chung
and Hui [16] have also used this method of normalisation in their
study of energy efficiency of office buildings in Hong Kong. The
normalised EPI and AAhEPI for each of the facilities for a sample of
‘n’ facilities, with ‘k’ significant operational characteristics are thus
given by