4
$\begingroup$

I am looking for some advice on how to take heart rate (HR) and electro-dermal activity (EDA) data files collected via an Empatica e4 and look for correlation between them.

My research project is looking at emotional responses to artworks in museums. I have individual participant data and what I am finding is that when visualised simply some emotional responses provoke EDA and other HR. What I would like to do is run an analysis to show whether there is a relationship between the two. I suspect not, and that this might be a pointless exercise, but it's something I want to have in my back pocket when questioned about these relationships and to use to defend my approach.

Has any one else done this, and if so, what was the best way?

$\endgroup$
3
$\begingroup$

HRV measures

Measures of heart-rate variability (HRV) are primarily calculated with the Inter-beat interval (IBI), also referred to as RR-intervals or NN-intervals (see this question). Some variables in the time domain are :

  • SDNN/SDANN
  • NN50 / pNN50
  • SDSD
  • rMSSD

And many more (see also the rHRV tutorial, which also describes frequency-domain measures).

EDA measures

Electrodermal activity can be quantified by calculating the galvanic skin level (GSL) or the galvanic skin responses (GSR; see also this question; Bouscein (2012)). GSL is a slow drifting tonic level of conductivity, whereas GSR are spikes in activity after some events. You can quantify these values by determining the average amplitude of the GSL/GSR’s within a specific time period, or calculating the amount of GSR’s.

Correlations

When you have acquired the variables you can start calculating correlations. You do have to ensure that the time intervals of these variables are the same. E.g. the average values of each variable are calculated over five-second windows. In turn, each five-second period can be considered as a data point.

Here are some results I found in a pilot-study I performed with two participants. Because of the little amount of data, not every correlation may be what you expect. pNN50, SDNN, LFHF, LF and HF are the HRV data. numGSR (amount of spikes), eda_p_filt (mean amplitude of GSR spikes) and eda_t (tonic skin activity) are the electrodermal-activity measures. The IWS is a subjective measure of workload (Kramer, Johnson and Zeilstra, 2016),

enter image description here


References

Bouscein, W., Roth, W. T., Dawson, M. E., & Filion, D. L. (2012). Publication recommendations for electrodermal measurements. Psychophysiology, 49, 1017-1034.

Garcıa, C. A., Otero, A., Vila, X., Méndez, A., Rodrıguez-Linares, L., & Lado, M. J. (2013). Getting started with RHRV.

Kramer, R., Johnson, A., & Zeilstra, M. P. (2016). The Integrated Workload Scale–Translation and validation of a subjective workload scale. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit.

What is the difference between RR-intervals and NN-intervals in HRV-data?

Open-source software for analyzing Electrodermal activity

$\endgroup$
  • $\begingroup$ @Jess I just came to think that I have a nice figure of correlations between these measures (and a subjective workload measure) in a small study I did. After my vacation I'll try to look it up :) $\endgroup$ – Robin Kramer Aug 15 '17 at 20:35
  • $\begingroup$ Brilliant. Thanks so much for this, Robin! I really appreciate it. $\endgroup$ – Jess Aug 16 '17 at 18:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.