Two models predict the same average energy use for a device under typical operation. One model includes many poorly measured parameters; the other uses a simpler balance with well-supported inputs. Which choice is best justified for estimating average use under those typical conditions?