Ranking with respect to the reference

Excellent	Good	Fair	Poor
SIVdata (1)	bdelouis (2)	holden (4)   sblindjb (5)   sblindjz (6)	festa (3)

Dissimilarity matrix

Solution	SIVdata	bdelouis	festa	holden	sblindjb	sblindjz
SIVdata	0.00	11.44	54.23	29.21	14.20	26.01
bdelouis	11.44	0.00	49.54	35.72	11.18	24.09
festa	54.23	49.54	0.00	82.75	76.04	81.92
holden	29.21	35.72	82.75	0.00	30.81	21.97
sblindjb	14.20	11.18	76.04	30.81	0.00	21.90
sblindjz	26.01	24.09	81.92	21.97	21.90	0.00

Each element represents a measure of dissimilarity between pairs of models obtained using normalized square metric

Note
This analysis employs multidimensional scaling (Borg and Groenen, 2005) to compare the slip models. The dissimilarities are obtained for all pairs of slip models, viz. dissimilarity matrix and embedded in low dimensional Euclidean space. We define four categories or levels of similarity: ”excellent”, ”good”, ”fair”, and ”poor” with respect to a reference (or target) solution or a centroid measure in absence of a reference solution.

Graphical interpretation for the similarity scale is as follows:

Excellent	Solution/s located inside the innermost circle
Good	Solution/s located between the innermost and middle circle
Fair	Solution/s located between the middle and outermost circle
Poor	Solutions/s located outside the outermost circle

For more details on the technique:

1.	Borg, I. and P. Groenen (2005). Modern Multidimensional Scaling, 2nd edition. New York, Springer
2.	Razafindrakoto, H. N., Mai, P. M., Genton, M. G., Zhang, L., and Thingbaijam, K. K. (2015). Quantifying variability in earthquake rupture models using multidimensional scaling: application to the 2011 Tohoku earthquake. Geophysical Journal International, 202(1), 17-40.