Table 6

Comparison of the investigated methods

lasso

adaptive lasso

PLS

Ridge Regression

shrinkage

properties

testing needed

no

no

yes

yes

yes

run-time

high

very high

medium

medium

low

simulation results

density

very dense

between lasso & non-sparse methods

very sparse

very sparse

very sparse

mean-squared-error

low

low

high

medium

medium

real-world-study

density

too dense on t.cell

too dense on t.cell

very dense

-

-

repeated measurements

too dense

too dense

dense

-

-

stability under resampling

medium

medium

medium

medium

good

Krämer et al. BMC Bioinformatics 2009 10:384   doi:10.1186/1471-2105-10-384