Abstract
In this paper we focus on the variable-selection peformance of solar on theempirical data with complicated dependence structures and, hence, severemulticollinearity and grouping effect issues. We choose the prostate cancerdata and the Sydney house price data and apply two lasso solvers, elastic netand solar on them (code can be found at \url{https://github.com/isaac2math/}).The results shows that (i) lasso is affected by the grouping effect andrandomly drop variables with high correlations, resulting unreliable anduninterpretable results; (ii) elastic net is more robust to grouping effect;however, it completely lose variable-selection sparsity when the dependencestructure of the data is complicated; (iii) solar demonstrates its superiorrobustness to complicated dependence structures and grouping effect, returningvariable-selection results with better stability and sparsity. Also, suchstability and sparsity make solar a reliable variable pre-estimation filter ofa linear dependence structure esimation (linear probablistic graph learning).The linear probablistic graph estimated on the variable selected by solarreturns an intuitive, sparse and stable dependence structure.