Abstract: In general,the data fusion modes of original measurements or derived observations,e.g.the partial estimated results,are identical in least squares principle.However,in the situation of incomplete functional or stochastic models,the data fusion modes of original measurements and derived observations are different.The fusion modes of the observations and derived observations with systemic errors or stochastic model errors are discussed,the properties of the two fusion modes are analyzed and compared in theory and illustrated by an actual example.