Theoretical assessment of the impact of noise on heuristic parameter inference methods for surface-confined non-catalytic voltammetry experiments

Abstract

• Comparison of inference methods for electron-transfer parameters using simulation. • We observe accuracy of estimates under varying quantities of synthetic noise. • Majority of accuracy loss comes from discarding total current during analysis. When analysing a voltammetric experiment, the goal is often to obtain an estimate of parameter values — as defined by models which predict the current obtained by applying a time-varying potential input to an electroactive species under investigation. There are a variety of methods used to obtain these estimates, such as the heuristic approach, which involves extracting some feature from one or multiple voltammetry experiments, and obtaining an estimate of one or two parameters by analysing the extracted feature(s). In this paper we simulate voltammetry data using pre-defined parameter values for a purely Faradaic surface-confined redox process, and then attempt to recover these values in the presence of varying amounts of synthetically-added noise, using both heuristic methods and a Bayesian inference approach. We show that the Bayesian approach recovers the true parameter values with either a greater or equivalent level of accuracy when compared to the heuristic approach, depending on the form of the data analysed. We show that the loss of accuracy for the heuristic case is primarily driven by the discarding of large amounts of data during the feature-extraction process.