This paper proposes a method to reproduce human hand surface shape precisely by combining statistical adjustment with skeletal subspace deformation (SSD). Plaster molds of the hand in variable postures were fabricated for one subject and their 3D shapes were obtained by computer tomography scan. A SSD model was prepared to compute adjustment vector. Feature points were manually selected from each plaster mold and SSD model. Difference of each feature point on a SSD model and corresponding one on a plaster model was collected from all the plasters and analyzed with postures to estimate adjustment vector for the feature point on SSD model in arbitrary posture. The rest of the points were adjusted using adjustment vectors of nearest three feature points.