We have recently implemented and tested the direct reconstruction of sinogram data to dense images of kinetic model parameters . In addition, we have recently applied our algorithms to brain data acquired with 18F-fallypride imaging of a monkey (2]. As a multi-dimensional parameter estimation exercise, direct reconstruction to parametric images can be thought of as generating thousands of model-fitted curves (the prediction of measured sinograms) simultaneously. Because the resulting parametric images are only as good as the fits to the data, one would like to have a means of evaluating the "goodness of fit" of each of the model-fitted curves. The size of the data set involved (4D PET data) presents unique problems in the visualization of the fits. In this paper, we propose measures to objectively evaluate the "goodness of fit" of the model to the PET sinograms in orders to evaluate the precision of the parametric images and the validity kinetic model. The techniques presented are, in part, extrapolations of standard parameter estimation techniques  to multi-dimensional estimates and are adapted to the tomography paradigm.