1: \begin{abstract}

2: 

3: In variable or graph selection problems, finding a right-sized model or controlling the number of false positives is notoriously difficult. Recently, a meta-algorithm called \stabilityselectionfirst \ was proposed that can provide reliable finite-sample control of the \fprpure. Its benefits were demonstrated when used in conjunction with the \lasso \ and orthogonal matching pursuit algorithms.

4: 

5: In this paper, we investigate the applicability of \stabilityselection \ to \algostructP: the \grouplasso \ and the \siolpure. We find that using \stabilityselection \ often increases the \accuracy \ of both algorithms, but that the presence of complex structure reduces the reliability of error control under \stabilityselection. We give strategies for setting \tuningP \ to obtain a good model size under \stabilityselection, and highlight its strengths and weaknesses compared to competing methods \screenandclean \ and \xval. We give guidelines about when to use which error control method.

6: 

7: %SS was proposed to improve power, theoretical robustness, reliable error control

8: %use it on SMR

9: %R1:consistently more power

10: %R2:error bound reliable

11: %R3:tradeoff

12: 

13: \end{abstract}

14: