SUPPLEMENTARY DATA
A separate file “P-values_number_of_VMM_components_Supplement1.xls” is provided
Supplement 1 A comparison of the range of p-values of the c2-tests when using different numbers of von Mises mixture (VMM) components. Nc – number of VMM components ranging from 1-5. AIC – Akaike information criterion, BIC – Bayesian information criterion. Two figures are shown per dataset and criterion in the format of a(b), where the second number b equals the Nc according to the AIC and BIC criterion and the corrected first number a equals the Nc remaining after removal of the weak and redundant components with probability<0.1 (too weak components; Figure 5)) or shape κ<0.75 (nearly uniform distributions). The p-values were classified into four categories at [0,0.95[, *[0.95,0.99[, **[0.99,0.9999[, and ***[0.9999, 1.0]. Using only one VMM component (Nc=1) was insufficient for the successful fitting of most of the samples. Considering that using two VMM components (Nc=2) resulted in the successful fitting of all but two samples, the secondary component was too weak or had a uniform distribution and was therefore removed (see values 1(2) in the table). The use of three VMM components (Nc=3) was successful in all samples, but the tertiary component proved to be essential (see values 3(3) in the table) in only 10 of the 82 samples. The use of further increasing numbers of VMM components (Nc=4 or 5) proved to be redundant. Although increased p-values were obtained in some cases, most of the third, fourth, and fifth components exhibited only low probabilities or were approximately uniform.
Note. The table can be read using the following example explained for the dataset No. 183_09 (third from top, highlighted with italics): The use of only one von Mises distribution (Nc=1) successfully fitted the measured data. The use of two von Mises distributions (Nc=2) also successfully fitted the data, and none of the distributions were found to be either weak or redundant. The use of three von Mises distributions (Nc=3) revealed that the third component was weak and could be dismissed without weakening the fitting. The use of four (Nc=3) or even five (Nc=5) von Mises distributions was not helpful in this case because the additional distributions were either uniform or their probabilities were significantly low.
A separate file “VSMC_orientation_area_angles_location_shape_Supplement2.xls” is provided
Supplement 2. The list of aortic samples grouped according to the vascular segments and age groups. The results of the angle analysis are provided as well as the location (μ) and shape (κ) of the two von Mises mixture (VMM) components with the same probability. The intima media thickness values are provided as well.
2