A three-way dynamic panel threshold regression model for change point detection in bioimpedance data

Bioimpedance refers to the measurement of the electrical impedance of biological tissue and it can be used to monitor its physio-chemical changes with important implications in a wide range of applications, including bioengineering and medical diagnostics. Recently bioimpedance analysis is gaining popularity in the food industry due to its ability to provide accurate and timely information about food quality, thereby reducing waste and improving product quality. Motivated by an original experiment on fruit bioimpedance, we develop a three-way panel dynamic threshold model and propose it as a method to detect the change point, i.e. fruit ripening time. We develop an estimation method based on the generalized method of moments and the instrumental variables estimator, and define a criterion to compute the change point. The performance of the proposed model is investigated through an extensive Monte Carlo simulation study in terms of the accuracy of both coefficients’ and thresholds’ estimates and the change point detected. The developed method is applied to an innovative fruit bioimpedance panel data set from two different electrical impedance analyzers, a bench-top device and a portable device. The method provides significant threshold and meaningful ripeness times. The new model is implemented and available through the R package PanelTM.