Robotic gait training has been introduced recently in rehabilitation, and the related researches have been conducted to propose more effective mechanism and control. To automate gait training, position control with reference trajectory has been adopted in many researches. However, there remain problems such as enhancing self-dominated gait or adapting to asymmetry or individual difference to apply robotic gait training to moderately affected hemiplegia patient. To solve this problem, we quantified the manual pelvic assistance (handling) provided by physical therapy, which can enhance patient's self dominated gait individually. In this paper, the physical model of handling was proposed based on the measurement and verified by multiple regression analysis.