In the development of medical rehabilitation tools, muscle signals or electromyographic (EMG) are often used as feedback from the body in order to facilitate normal movement patterns. Through this signal, the force, torque, and angle can be estimated to produce the desired movement in the robot to assist patients in rehabilitation therapy. There had been various research done on estimating the extremities' joint angle, including the system identification method with Auto-Regressive Integrated Moving Average with Exogenous Input (ARIMAX) model structure. This research aims to model biceps brachii's EMG signal and the elbow joint angle using the ARIMAX model structure. In modeling, the order of the model was selected using the methods of minimizing the AIC (Akaike Information Criteria) and RMSE (Root Mean Square Error) values. The reported results showed the comparison of the outcome and model orders from three data sets with a frequency sampling of 47.6690 Hz, l66.3248 Hz, and 935.6743 Hz, respectively. The frequency sampling of l66.3248 Hz results showed the best accuracy performance of the ARIMAX model structure. The model order which is selected based on the minimization of the RMSE value obtained the best estimate, while the order models were selected based on the minimization of the AIC value showed a more consistent performance on the data estimation and validation.