Statistical Analysis of SCR Data

We chose the statistical approach for the SCR data based on the hypotheses we developed before we started testing our subjects. The SCR we collected in LabChart was merged with the PsychoPy files for each subject to synchronize the events with the physiological data.

We downsampled the SCR from 1,000 to 100 hertz (Hz) using Matlab (MathWorks^®, Aachen, Germany) for better computing performance. Afterward, we used Ledalab² for the analysis of our SCR data. We used the approach Benedek and Kaernbach (2010a, b) described to analyze our SCR data. They used a continuous decomposition analysis (CDA) which separates the SCR data into continuous signals of phasic and tonic activity. The tonic activity reflects the slow varying SC level when no stimulus is applied. It can be estimated by measuring SC data of the participants without them being stimulated as a baseline. The phasic activity is the vast-varying activity with a zero baseline and event-related changes that portray responses specific to a stimulus and non-specific response. We then conducted an event-related analysis of the phasic activity by inserting the time stamps of the Tetris game start – each start of a Tetris game was an individual event.

The statistical analysis was performed using SPSS 25 (IBM, Ehningen, Germany). As a dependent variable we used the mean SC value of the phasic driver within the response window (global mean) in a 10 s time interval starting 0.8 s after the start of the odor application for each round. We tested for an impact of the cognitive load setting (low, high) on the global mean value using a LMM with the fixed factor cognitive load (low, high) and a random factor for subject. The threshold for this test was set to p < 0.05.