Current design process for orthopedic implants relies on limited knowledge about shape and mechanical properties of the bones of interest. Nowadays implant design is based on reduced information deriving from literature (e.g. bone angles and lengths) and does not take into account differences among populations (ethnicity, height, gender ...). For these reasons we propose a method to model bone shape variations for different populations. The proposed model is based on recent advances in image processing to create compact mathematical representations of bone anatomy. About 140 CT images were collected. These images were segmented and registered to find the anatomical correspondences among them. Based on this registration, a statistical shape model was built to compute the average bone, variations of bone shape across the population and to generate new bone instances. The finite element method was then used to calculate the mechanical performance of the generated instances. The developed model was used to compare bone stiffness between male and female populations. Results showed no significant difference between male and female bone stiffness. This was affected by the fact that all instance had the same dimensions and the same mechanical properties. The combination of statistical shape model and finite element calculation has the potential to become a powerful tool to assess orthopedic procedures and implant design.