At Step 602, the grid of a wafer loaded into an exposure apparatus is approximated by a mathematical function fitting up to, for example, a cubic function, and at Step 612, the magnitude of a residual error between the position of a sample shot area obtained by the function and an actually measured position is compared with a predetermined threshold value. GCM measurement is performed in a mathematical function mode in a subroutine 616, or it is performed in a map mode in a subroutine 616, on the basis of the result of the comparison. In addition, it is determined whether to extract non-linear components from the wafer in each lot on the basis of a variation in the temperature of the wafer (Step 622) and a variation in random error between the wafers (Step 624).