Spatial prediction methods are briefly reviewed in an attempt to apply the methodology to a manufacturing problem of geometric shape estimation. While stochastic kriging models and deterministic spline models differ in origin and professed objectives, the latter may be viewed in terms of the former; in particular in terms of intrinsically stationary isotropic kriging. Several methods such as surface splines and multiquadrics are interpreted in kriging terms to the effect that the methodology is looked upon as being unified under the name of spatial prediction. The main purpose of this work is to develop an approach to the estimation of form error of machined parts or wear-pattern of used parts, presumably measured by 3-coordinate measuring machine, in an effort to have the analysis rooted in spatial prediction. While this study may find applicability in machining generally, the estimation of geometric objects featuring roundness is focused on.