Industrial manipulators require high precision and accuracy to perform repetitive tasks without affecting the robotic platform and the different agents interacting in the operating environment. Therefore, the end-effector position is important for accurate task execution. Machine learning and statistical methods using kinematic models to estimate the end-effector position have significant limitations in accuracy, robustness, and generating capacity in the face of dynamic or unmodeled conditions. In this context, a comparative analysis between a Bayesian regression model and a maximum likelihood-based model is proposed to estimate the end-effector position from kinematic information of the manipulator, considering a priori knowledge about the effector positions and joint angles. Error metrics such as the mean square error, coefficient of determination, and generalization capacity in untrained trajectories are used to evaluate the performance of the models. The results disclosed that the maximum likelihood model exhibits greater computational efficiency under ideal conditions. In contrast, Bayesian models demonstrate a superior capacity to represent the uncertainty inherent in the kinematic model.