This paper presents a methodology to derive mathematical models of active power components based on statistical theory and datasheet parameter extraction. The models have the objective of providing reliable data of gate-to-drain voltage (VGD), gate-to-source voltage (VGS) and direct current (ID) for nominal operating temperature range using only datasheet points. For achieving that, statistical learning theory is used to fit models in a reliable and systematic way. Besides predicting accurately static operation points, the proposed model can be used to estimate dynamics and switching losses in power switches. The model is validated using static and dynamic data available in a datasheet. The model is used to analyze a MOSFET behavior in typical power electronics applications. Finally, it is discussed the data availability in datasheets in order to easily create accurate models.