Evaluation of CDA and CvxEDA Models

Abstract

There is no universally accepted experimental protocol for the validation of EDA analysis algorithms. A characterization in terms of sensitivity and specificity, as typically done in a pattern recognition framework, is not directly applicable in this context since there is no one-to-one correspondence between external stimuli supposed to elicit ANS responses and skin conductance responses (see Bach and Friston, Psychophysiology 50(1):15–22, 2013 for a discussion). Unless the sympathetic nerve activity is also recorded through microneurography, failure to detect a phasic SC response after the occurrence of an experimental stimulus may be equally ascribed to a low sensitivity of the algorithm under study or, alternatively, to the inability of the stimulus to consistently elicit a phasic response. Similarly, detection of phasic activity in the absence of stimulation may be caused by electrodermal changes that are not stimulus-elicited but spontaneous and non-specific, possibly a result of muscular contractions or respiratory irregularities (Boucsein, Electrodermal activity, 2nd edn. Springer Science & Business Media, New York, 2012).