Dataset Information

Group SLOPE - adaptive selection of groups of predictors.

ABSTRACT: Sorted L-One Penalized Estimation (SLOPE, Bogdan et al., 2013, 2015) is a relatively new convex optimization procedure which allows for adaptive selection of regressors under sparse high dimensional designs. Here we extend the idea of SLOPE to deal with the situation when one aims at selecting whole groups of explanatory variables instead of single regressors. Such groups can be formed by clustering strongly correlated predictors or groups of dummy variables corresponding to different levels of the same qualitative predictor. We formulate the respective convex optimization problem, gSLOPE (group SLOPE), and propose an efficient algorithm for its solution. We also define a notion of the group false discovery rate (gFDR) and provide a choice of the sequence of tuning parameters for gSLOPE so that gFDR is provably controlled at a prespecified level if the groups of variables are orthogonal to each other. Moreover, we prove that the resulting procedure adapts to unknown sparsity and is asymptotically minimax with respect to the estimation of the proportions of variance of the response variable explained by regressors from different groups. We also provide a method for the choice of the regularizing sequence when variables in different groups are not orthogonal but statistically independent and illustrate its good properties with computer simulations. Finally, we illustrate the advantages of gSLOPE in the context of Genome Wide Association Studies. R package grpSLOPE with an implementation of our method is available on CRAN.

SUBMITTER: Brzyski D

PROVIDER: S-EPMC6583898 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Group SLOPE - adaptive selection of groups of predictors.

Brzyski Damian D Gossmann Alexej A Su Weijie W Bogdan Małgorzata M

Journal of the American Statistical Association 20180806 525

Sorted L-One Penalized Estimation (SLOPE, Bogdan et al., 2013, 2015) is a relatively new convex optimization procedure which allows for adaptive selection of regressors under sparse high dimensional designs. Here we extend the idea of SLOPE to deal with the situation when one aims at selecting whole groups of explanatory variables instead of single regressors. Such groups can be formed by clustering strongly correlated predictors or groups of dummy variables corresponding to different levels of ...[more]

PMID: 31217649

Similar Datasets

Project description:Mixing gestating sows implies hierarchy formation and has detrimental consequences on welfare. The effects of social stress on the most vulnerable individuals may be underestimated and it is therefore important to evaluate welfare between individuals within groups. This study aimed at investigating the impact of social status and previous experience in the group on well-being of sows housed in large semi-static groups. We assessed aggression (d0 (mixing), d2, d27, d29), body lesions (d1, d26, d84) and feeding order on 20 groups of 46-91 animals. Social status was based on the proportion of fights won during a 6-hr observation period between d0 and d2. Dominants (29%) were those who won more fights than they lost, Subdominants (25%) won fewer fights than they lost, Losers (23%) never won any fight in which they were involved while Avoiders (23%) were never involved in fights. Resident sows (70%) were already present in the group in the previous gestation while New sows (30%) were newly introduced at mixing. Subdominants and Dominants were highly involved in fights around mixing but this was more detrimental for Subdominants than Dominants, Losers and Avoiders since they had the highest body lesion scores at mixing. Avoiders received less non-reciprocal agonistic acts than Losers on d2 (P = 0.0001) and had the lowest body lesion scores after mixing. However, Avoiders and Losers were more at risk in the long-term since they had the highest body lesions scores at d26 and d84. They were followed by Subdominants and then Dominants. New sows fought more (P<0.0001), tended to be involved in longer fights (P = 0.075) around mixing and had more body lesions throughout gestation than Resident sows. Feeding order from one-month post-mixing was influenced both by the previous experience in the group and social status (P<0.0001). New sows, especially with a low social status, are more vulnerable throughout gestation and could serve as indicators of non-optimal conditions.

Dataset Information

Group SLOPE - adaptive selection of groups of predictors.

Publications

Group SLOPE - adaptive selection of groups of predictors.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets