Amethod to evaluate statistical estimation error of the mean of the local average of a geotechnical parameter is proposed and verified in this paper. The soil profile is model by a stationary random field, which is simplification and idealization of the real ground. The local average at an arbitrary point in the random field is estimated from limited samples obtained from the same filed. The two estimation methods are proposed, namely the General estimation and the Local estimation. The relative position of samplings and of the structures are not considered in the former, whereas they are considered in the latter. The theory is merely idealization and simplification of the real problem. The theory is verified based on intensive cone penetration tests (CPT) at three sites. The method is basically cross verification which compares the estimated estimation variance with calculated true values based on the intensive CPT measurements.