Table 4. Ridge-regression vs lasso.

par.	obs. stat.	p-value
sample size ()	28.003	<0.001
number of features ()	99.312	<0.001
saturation ()	78.062	<0.001
signal-to-noise ()	0.949	0.541
correlation ()	85.745	<0.001
sample size vs number of features ()	3.948	<0.001
sample size vs saturation ()	8.321	<0.001
sample size vs signal-to-noise ()	1.002	0.482
sample size vs correlation ()	0.996	0.490
number of features vs saturation ()	41.886	<0.001
number of features vs signal-to-noise ()	1.023	0.426
number of features vs correlation ()	1.393	0.017
saturation vs signal-to-noise ()	1.017	0.443
saturation vs correlation ()	0.570	1.000
signal-to-noise vs correlation ()	0.695	0.986

Permutation tests for equality of the group distributions using distance components analysis (lines 2 to 6), and permutation F-tests for the presence of 2-by-2 interactions (lines 7 to 16), in the comparison of ridge-regression vs lasso. Results based on 999 permutations.