The effective modeling and predicting of respiratory motion in abdominal organs is crucial to the task of tumor treatments. Current approaches in statistical respiration modeling either build a subject-specific model which is only suitable for that single subject or create a population model by assuming a coherent population with a relatively simple distribution and, therefore, fail to account for variations of the breathing pattern among different subjects. To bridge this gap, we propose a more flexible method based on exemplar models, which is able to cope with heterogeneous population data and can be better adapted to a previously unseen subject. We have showed that, in contrary to principal component analysis based models, our method is capable of effectively utilizing complementary information provided by increasing number of examples taken from a population. In addition to being more robust against outliers, the proposed method also achieved lower mean errors in leave-one-out experiments.