An algorithm is presented for updating finite element models based upon a minimization of dynamic residuals. The dynamic residual of interest is the force unbalance in the homogeneous form of the equations of motion arising from errors in the model's mass and stiffness when evaluated with the identified modal parameters. The present algorithm is a modification and extension of a previously-developed Sensitivity-Based Element-By-Element (SB-EBE) method for damage detection and finite element model updating. In the present algorithm, SB-EBE has been generalized to minimize a dynamic displacement residual quantity, which is shown to improve test-analysis mode correspondence. Furthermore, the algorithm has been modified to include Bayesian estimation concepts, and the underlying nonlinear optimization problem has been consistently linearized to improve the convergence properties. The resulting algorithm is demonstrated via numerical and experimental examples to be an efficient and robust method for both localizing model errors and estimating physical parameters.