Improved statistical procedure for the evaluation of solar radiation estimation models

The t-statistic is proposed as a statistical indicator for the evaluation and comparison of solar radiation estimation models. A relationship was developed for the t-statistic as a function of the widely used root mean square and mean bias errors. Using data published in the solar energy literature, it was demonstrated that the use of the root mean square and mean bias errors separately can lead to the incorrect selection of the best model from a suite of candidate models. It is recommended that the t-statistic be used in conjunction with the root mean square and mean bias errors to more reliably assess a model's performance. The t-statistic has the added advantage of enabling the model tester to determine whether or not a model's estimates are statistically significant at a particular confidence level.