The Short Answer is… Yes. The ENERGY STAR score is a statistical measure and every statistical measure has some uncertainty. In order to give your property an ENERGY STAR score or tell you if you’re better than the national median, we need to be able to compute that national median value. For example, if you have a library, then we need to have statistics that tell us something about the entire population of libraries. When we compute these statistics, we use data from a representative sample of libraries to say something about the entire population of libraries. By definition, this process of extrapolating from sample data has some uncertainty.
The Longer Answer is… Because the ENERGY STAR score is more complicated than a metric like the mean or the median, there is no single, standard approach for estimating the uncertainty associated with it. Possible sources of uncertainty in calculation of the ENERGY STAR score include the regression model that we use to normalize for factors like hours of operation, and the process we use to understand the distribution (or, range) of energy use across all buildings of a given type. While it might be possible to estimate component uncertainty for these individual steps, there is no standard approach for combining those uncertainties.
EPA strives to limit uncertainty in the ENERGY STAR score. For example, one approach to limiting uncertainty is to avoid “over-fit” in the regression model. The regression allows us to normalize for key operational factors like hours and workers. In some cases, adding more factors can make the model appear better to have a better “fit” (i.e., result in a higher “R-Squared”), but increase uncertainty. Therefore, EPA focuses on the relatively small set of operational factors that are most important to enabling equitable comparisons among buildings. We welcome input on other approaches to limiting uncertainty while maintaining the value of the ENERGY STAR score.