This article presents an uncertainty analysis method for the 16-term error model, and applies the methodology in determining the best calibration standard combinations and their associated parasitic parameters. This article proposes a sensitivity analysis method using the complex-valued matrix differentiation to tackle the complexity of the 16-term error network, and the uncertainty propagation is further derived by the properties of the covariance matrix. A novel methodology is additionally proposed for calibration standard selection by combining the uncertainty analysis method with a parasitic-parameter extraction approach. The newly proposed methodology jointly determines the best combinations of calibration standards and their corresponding parasitic parameters and enhances the calibration accuracy and reliability of the 16-term error model. The effectiveness of the technique is validated by on-wafer measurements.