An error compensation method includes outputting at least one of a plurality of calibration-master conditions including a type of a calibration master that includes targets, obtaining measurement values of positions of the targets by detecting the positions of the plurality of targets under the calibration-master condition using a sensor mounted to the main spindle, and calculating an error value using the measurement value and a calibration value of the position of the target. The error compensation method further includes approximating a relation between the measurement values and the error values by a curve and a straight line, calculating a compensation parameter of a positioning error of a translational axis based on an approximate curve in a partial range of a stroke of the translational axis, and calculating the compensation parameter of the positioning error based on an approximate straight line in another range of the stroke of the translational axis.