Comment on Article by Monni and Tadesse

The subsection entitled Performance in the presence of high collinearity on page 15 starts with the sentence “It is reasonable to assume that the presence of highly correlated covariates may complicate the identification of relevant predictors.” Two new covariates, which are linear combinations of existing covariates, are then added to simulated data which had N = 50 samples and p = 200 covariates, with the conclusion that “. . . the algorithm is reasonably resistant to colinearity between predictors.” It’s not clear what’s intended here, since the dimensions of the data imply multicolinearlity, and the method is supposed to address cases in which the number of regressors is much larger than the sample size (e.g. page 2). Questions of interest with regard to multicolinearity would include: