Conditional variable importance for random forests

Strobl, Carolin; Boulesteix, Anne‐Laure; Kneib, Thomas; Augustin, Thomas; Zeileis, Achim

doi:10.1186/1471-2105-9-307

Cited by 2,599 publications

(2,159 citation statements)

References 29 publications

Supporting

Mentioning

2,041

Contrasting

Unclassified

Order By: Relevance

“…Later, Strobl et al [40] pointed out another issue of RF variable importance which shows a bias towards correlated predictor variables. The issue of correlated feature variables happens commonly in high-dimensional bioinformatics tasks, e.g.…”

Section: Revised Rf Feature Importancementioning

confidence: 99%

“…genomics. This paper [40] developed a conditional permutation scheme which used the partition automatically provided by the fitted model as a conditioning grid. The resulting measure was claimed to reflect the true impact of each predictor (variable) better than the original, marginal approach.…”

Section: Revised Rf Feature Importancementioning

confidence: 99%

“…Gene-gene interactions are importance factors to consider when selecting features for disease classification, however, popular univariate selection methods could not take them into account. Thus, researchers have proposed a number of techniques to capture the correlations between genes using random forest based variable importance [15,46,40,2]. Several related methods have been covered in Section 2.2.…”

Section: Analysis Of Microarray Gene Expression Datamentioning

confidence: 99%

See 2 more Smart Citations

Random Forest for Bioinformatics

2012

Ensemble Machine Learning

587

344

View full text Add to dashboard Cite

Section: Revised Rf Feature Importancementioning

confidence: 99%

Section: Revised Rf Feature Importancementioning

confidence: 99%

Section: Analysis Of Microarray Gene Expression Datamentioning

confidence: 99%

See 1 more Smart Citation

Random Forest for Bioinformatics

2012

Ensemble Machine Learning

587

344

View full text Add to dashboard Cite

“…Strobl et al [2008] suggested using a conditional permutation scheme to calculate VI. The variables correlated with the variable of interest are empirically determined, and then the partitions in the individual tree are utilized to permute the variable of interest within blocks.…”

Section: Other Variable Importance Measuresmentioning

confidence: 99%

Random Forests for Genetic Association Studies

Goldstein

Polley²,

Briggs

2011

Statistical Applications in Genetics and Molecular Biology

212

172

View full text Add to dashboard Cite

The Random Forests (RF) algorithm has become a commonly used machine learning algorithm for genetic association studies. It is well suited for genetic applications since it is both computationally efficient and models genetic causal mechanisms well. With its growing ubiquity, there has been inconsistent and less than optimal use of RF in the literature. The purpose of this review is to breakdown the theoretical and statistical basis of RF so that practitioners are able to apply it in their work. An emphasis is placed on showing how the various components contribute to bias and variance, as well as discussing variable importance measures. Applications specific to genetic studies are highlighted. To provide context, RF is compared to other commonly used machine learning algorithms.

show abstract

“…Here, the environmental variables are strongly covariant and the model contains more than one variable type (all continuous except for morphotype, which is categorical). To account for this and aid in interpretation of the rankings, an 170 unbiased, conditional variable importance ranking method was incorporated via the party package in R, which disentangles the most important variable from the model (Strobl et al, 2008). This method examines whether a correlation between the response variable and a predictor is conditional on another variable proceeding it in the tree, thereby identifying the most influential variable and demoting others (Strobl et al, 2008).…”

Section: Methodsmentioning

confidence: 99%

Factors Influencing Porosity in Planktonic Foraminifera

Burke¹,

Renema²,

Henehan³

et al. 2018

Preprint

View full text Add to dashboard Cite

Abstract. The clustering of mitochondria near pores in the test walls of foraminifera suggests that these perforations play a critical role in metabolic gas exchange. As such, pore measurements could provide a novel means of tracking changes in metabolic rate in the fossil record. However, in planktonic foraminifera, variation in pore size, density, and porosity have been 20 variously attributed to environmental, biological, and taxonomic drivers, complicating such an interpretation. Here we examine the environmental, biological, and evolutionary determinants of porosity in 718 individuals representing 17 morphospecies of planktonic foraminifera from 6 core tops in the North Atlantic. Using random forest models, we find that porosity is primarily correlated to size and habitat temperature, two key factors in determining metabolic rates. In order to test if this correlation arose spuriously through the association of cryptic species with distinct biomes, we cultured Globigerinoides ruber in three 25 different temperature conditions, and found that porosity increased with temperature. Crucially, these results show that porosity can be plastic: changing in response to environmental drivers within the lifetime of an individual foraminifer. This demonstrates the potential of porosity as a proxy for foraminiferal metabolic rates, with significance for interpreting geochemical data and the physiology of foraminifera in non-analog environments. It also highlights the importance of phenotypic plasticity (i.e., ecophenotypy) in accounting for some aspects of morphological variation in the modern and fossil 30 record.

show abstract

Conditional variable importance for random forests

Cited by 2,599 publications

References 29 publications

Random Forest for Bioinformatics

Random Forest for Bioinformatics

Random Forests for Genetic Association Studies

Factors Influencing Porosity in Planktonic Foraminifera

Contact Info

Product

Resources

About