Knowledge about metabolisable energy (ME) intake is crucial for various experimental settings in rodent studies. ME considers faecal and renal energy losses. In particular, faecal energy excretion can vary considerably between differentially composed diets. Thus determination of faecal energy losses, i.e. apparent energy digestibility, is the most important experimental approach to determine ME. Predictive equations for ME such as Atwater factors or an equation for pigs, which are frequently employed for rodent feed, consider an average energy digestibility for nutrients and average renal losses for protein. Both equations, however, were never validated for rat feed. We therefore determined experimentally the digestibility of energy (experimentally determined digestible energy − 5·2 kJ/g digestible protein) and nutrients of eleven natural and five purified rat diets and compared the present results with the predicted values. Compared with natural diets, digestibility of gross energy (GE) and nutrients was higher by about 20 % in the purified diets (P < 0·0001). Mean GE digestibility in natural diets amounted to 71·4 % (range 53·3–83·5 %; n 11). Atwater factors predicted ME with satisfactory accuracy in purified diets. In contrast, for natural diets, only the equation for pig feed gave acceptable estimates of ME. Choosing an inappropriate predictive equation for ME resulted in considerable error. For prediction of ME in mixed rat feed, we propose to use the equation for pig feed for natural diets and Atwater factors for purified diets. If the equation for pig feed cannot be applied we suggest using the lower modified Atwater factors instead of the ‘original’ Atwater factors to estimate the ME of a diet.