Skip to main content
. 2011 May 23;11:137. doi: 10.1186/1471-2148-11-137

Table 2.

Performance of alternative models for the evolution of mandible shape using principal components.

Models with 3 PCs
BM OU.2 OU.3 OU.4 OU.5

AICc -503.15 -521.09 -557.23 -571.23 -568.09
Δ(AICc) 68.08 50.14 14 0 3.14
W(AICc) 0.0000 0.0000 0.0008 0.8272 0.1721
SIC -477.55 -472.61 -501.83 -509.29 -500.05
Δ(SIC) 31.74 36.68 7.46 0 9.24
w(SIC) 0.0000 0.0000 0.0232 0.9673 0.0095
DOF 9 18 21 24 27

Models with 5 PCs

BM OU.2 OU.3 OU.4 OU.5

AICc -953.13 -963.13 -994.16 -1016.73 -1022.49
Δ(AICc) 69.36 59.36 28.33 5.76 0
w(AICc) 0.0000 0.0000 0.0000 0.0532 0.9468
SIC -886.85 -839.13 -857.41 -867.95 -862.51
Δ(SIC) 0 47.72 29.44 18.9 24.34
W(SIC) 0.9999 0.0000 0.0000 0.0001 0.0000
DOF 20 40 45 50 55

Model names defined as follows: BM - Brownian motion, OU - Ornstein-Uhlenbeck with two (OU.2), three (OU.3), four (OU.4), and five (OU.5) adaptive optima (see text and Fig. 7 for details). AICc is the corrected Akaike Information Criterion for small sample sizes, Δ(AICc) is the difference relative to the model with smaller AICc, w(AICc) is the Akaike weight for each model. SIC is the Schwartz Information Criterion, Δ(SIC) and w(SIC) are the differences and weights as defined above. DOF are the numbers of parameters estimated by each model. All models were multivariate. The upper part of the table shows models fitting the first three shape PCs simultaneously, whereas the bottom part of the table shows models fitting the first five shape PCs.