The use of calibration-free or minimal calibration chemometric modeling approaches has garnered increased interest as the application of process analytical technologies (PAT) has expanded in the pharmaceutical industry. Pure component models such as iterative optimization technology (IOT) algorithms require only pure component spectra and no mixture samples as model inputs. However, the deployment of pure component models for PAT is limited by the lack of model diagnostics to establish confidence in model predictions and identify outliers. In the presented research, a novel diagnostic for use with IOT, called NAS-T, is proposed based on comparing the shape of net analyte signals (NAS) between the pure components and a sample mixture. Circular statistics were used in the analysis of NAS-T. The new diagnostic was applied in parallel with the partial least squares (PLS) Hotelling's T2 diagnostic to identify outlier near-infrared spectral samples from a set of pharmaceutical powder mixtures. The outliers detected by the NAS-T diagnostic included both chemical interferences (contamination) and physical interferences (process variations). Preprocessing the mixture sample spectra produced trends in the model diagnostics that were consistent between IOT and PLS. These results support the adoption of IOT and other pure component models for PAT applications in the pharmaceutical industry.