The aim of this dissertation is to develop a methodology for parameter identification in transmission loss modeling at the system level. Current studies indicate that potential uncertain parameters in the individual analytical formulas can lead to large prediction errors in the overall transmission loss simulation. It should be asked: how to perform experiments for efficient and informative characterization of transmission total loss behaviors, and how to deal with parametric uncertainties for minimizing their unnecessary correlations? This dissertation proposes an adaptive measurement strategy that ensures an iterative evaluation of surrogate models to explore new measurement candidates in the region of high interest. In contrast to the traditional factorial design, the adaptive strategy benefits from reducing the total number of measurement points by about half while keeping the same mapping quality. With the measured total losses, an inverse analysis method is developed for dealing with parametric uncertainties in the analytical simulation. Sensitivity analysis is first employed not only to determine the degree of importance of all uncertain parameters but also to minimize their unnecessary correlations in subsequent identification processes. Moreover, stochastic identification is applied here to avoid possible ill-conditioned inverse problems. With the help of the developed methodology, two repeated inverse analyses reveal stable identified parameters and improved simulation results. Following this, a detailed study of the influences of all operating conditions on total losses and individual loss proportions is performed, which unfolds the value of the methodology in the application of transmission efficiency investigations.