In order to enable optimization of the kinemetic calibration condition of a robot considering its task, we proposed an estimation method of output pose error after calibration and the error evaluation index. The index is derived from linearization of relationship between output pose error and measurement error. It depends on kinematic parameters of the robot, calibration condition and error evaluation condition specified by the task. Simulations considering measurement error for the Stewart platform, one of the six-dof parallel robots, were conducted to clarify the relationship between the proposed index and positioning-error after calibration. Effectiveness of the calibration condition obtained by the proposed index was confirmed through comparison with conventional method. Results of experiments using a Stewart platform were shown to support the proposed method.