Supplement

To classify RR sequences,
we explored linear discriminant analysis (LDA) commonly used for detection of
stress (Melillo et al., 2011; Kaur
et al., 2014). We used a
log-transformation to meet assumptions of homogeneity of variance and
normally-distributed residuals. The homogeneity of
variances was tested by means of Cochran C statistic. In order to evaluate the overall performance of the classifiers, results
of the rest-stress comparisons were quantified by the criteria of sensitivity,
specificity, precision, and accuracy. The highest estimates of
the total classification accuracy and sensitivity and was achieved by α1 and by
recurrent plot measures (Supplement 2).