Computational techniques for representing and analyzing full-wafer metrology data are developed for chemical vapor deposition and other thin-film processing applications. Spatially resolved measurement data are used to produce "virtual wafers" that are subsequently used to create response surface models for predicting the full-wafer thickness, composition, or any other property profile as a function of processing parameters. Statistical analysis tools are developed to assess model prediction accuracy and to compare the relative accuracies of different models created from the same wafer data set. Examples illustrating the use of these techniques for film property uniformity optimization and for creating intentional film-property spatial gradients for combinatorial CVD applications are presented. (C) 2009 Elsevier B.V. All rights reserved.