Table 2.

Statistical models fitted to untransformed data for scaling 2D (mm) to 4D (mm) in the right hand. AIC, Akaike's information criterion; ΔAIC, Akaike difference.

model	AIC	ΔAIC	inference
Straight line, no intercept, with lognormal heteroscedastic error	1984.1	61.0	no empirical support
Straight line, no intercept, with normal, heteroscedastic error	1983.7	60.6	no empirical support
Failed to converge. Convergence criterion changed to 0.011
Straight line, no intercept, with normal, homoscedastic error	1979.9	56.8	no empirical support
Three-parameter power function with normal, heteroscedastic error	1929.0	5.9	plausible alternative
Failed to converge. Convergence criterion changed to 0.014
Two-parameter power function with normal, heteroscedastic error	1928.8	5.7	plausible alternative
Failed to converge. Convergence criterion changed to 0.013
Straight line, intercept, with lognormal heteroscedastic error	1928.1	5.1	plausible alternative
Straight line, intercept, with normal, heteroscedastic error	1927.3	4.3	plausible alternative
Failed to converge. Convergence criterion changed to 0.01
Three-parameter power function with lognormal, heteroscedastic error	1926.5	3.5	plausible alternative
Failed to converge. Equation rearranged and converged
Two-parameter power function with lognormal, heteroscedastic error	1925.9	2.8	plausible alternative
Straight line, intercept, with normal, homoscedastic error	1924.6	1.6	essentially equivalent
Three-parameter power function with normal, homoscedastic error	1923.8	0.8	essentially equivalent
Failed to converge. Equation rearranged and converged
Two-parameter power function with normal, homoscedastic error	1923.1	0	best