Polarization curves are measurements of net current versus potential for an electrochemical system. The measured current is the sum of currents from several anodic (+) and cathodic (–) reactions, which can not be measured independently. It is hoped that chemometric models can be used to determine important characteristics of the underlying reactions. Several difficulties due to the nature of polarization data have been encountered. The domain of potentials for which the current can be measured depends on the variables under study, leading to non-random missing values at the extreme values of potential. The uncertainty in the measured current depends on the current, not the potential. In this work, missing values have been replaced by their estimates as the model is being generated, and iteratively re-weighted least squares has been used to overcome the latter problem. There is also noise in the potential (independent variable), including random missing values. A Savitzky-Golay filter has been used to interpolate the missing values and smooth the data prior to analysis.