Model based Approaches
The main issue with the conventional peak detection method for analysis described so far is how it handles superimposed SCRs [5]. In this case, an SCR onset starts before the termination of the previous SCR, e.g., when the ISI is very short (e.g., 2 seconds). Based on the time that the second SCR starts, there might not be enough time for the first SCR to recover or, in some cases, even to reach its peak. This makes the conventional analysis challenging, because we cannot detect peaks and measure recovery time of an incomplete SCR. To address this issue, researchers have proposed model-based approaches. These models automatically decompose the EDA signal and measure phasic and tonic parameters. These approaches assume that sympathetic processes that generate EDA can be described as time series of a mathematical model (e.g. time-invariant model [3]) and then use the recorded EDA data to estimate the time series process that generated the EDA values. With this we can understand central process caused an EDA data and then we can measure EDA parameters.
One of the most well-known methods for decomposing EDA data to tonic and phasic drivers is the Continuous Deconvolution Analysis (CDA) [5, 6], which uses a mathematical model of sweat diffusion. CDA is implemented in the LedaLab toolbox. Greco et al. [14] proposed cvxEDA, which considers EDA as an autonomic nervous system activity and models it using convex optimisation. Implementations of this model are publicly available1. Bach et al. [3] proposed another approach that assumes a linear time invariant model generates SCRs. This model is implemented in PsPM toolbox. SparsEDA [16] is another publicly available2 model which is fast and efficient in lower sample rates and suitable for wearable devices. In addition to above-mentioned models, many other models proposed for psycho-physiological modeling each assuming a different type of function is driver of SCR. Reviewing all these methods is out of scope of this paper; however, to find the best model to use for your study you can refer to reviews [2, 4, 15, 19] on these methods.