OBJECTIVES: Poor preservation of hominin pelvises and the lack of soft tissue in the fossil record inhibits researchers' abilities to ascertain the true geometry of hominin pelvic girdles. The reconstruction process becomes subjective, largely relying on researchers' anatomical expertise, particularly, when the sacrum is absent or cannot be used to orient the hip bones. The bilateral symmetry of the pelvis, however, offers an opportunity to use one side to reconstruct potentially missing data on the other side. METHODS: We developed a regression model to predict the translation and rotation actions that are needed to transform a hip bone onto the location of its pair. We collected landmarks and curve semilandmarks on a training sample of medical CT scans of 103 adult humans. A reduced rank regression model was trained to predict the values that would fit each right hip bone on its left pair. Then, we applied the model to two reconstructions of the Kebara 2 Neanderthal pelvis and assessed how well, it predicted the reconstructions (assuming the sacrum was absent), which were made using the preserved sacrum. RESULTS: Euclidean errors from the model were significantly lower than errors from a mean form model and an observed form pairwise model. CONCLUSION: Regression modeling that takes advantage of bilateral symmetry can be used to reliably predict "missing" human hip bones and Kebara 2's reconstructed left hip bones. This method can be employed in conjunction with a researcher's anatomical expertise and other techniques to reduce subjectivity in the fossil pelvis reconstruction process.