Description Usage Arguments Value Author(s) References Examples

The workhorse function - estimates statistical significance of feature importance by permuting the response variable

1 2 3 |

`response` |
a character vector or a factor for classification containing the group memberships for classification, a numeric vector for regression |

`predictors` |
A matrix consisting of features (measurements) corresponding to samples. The orientation per se does not matter - the function orients them correctly for Random Forest learning. |

`n.perms` |
Number of permutations to estimate significance. If the number of all possible permutations is less than this the latter will be used for estimation. |

`alpha` |
The significance level threshold of p.values for estimating false discovery rate using the two-step BH method for correlated test statistics, as implemented in the multtest package's mt.rawp2adjp function. |

`mtry` |
see ?randomForest for details - defines how many features are randomly sampled for building trees |

`type` |
string, set to "classification" or "regression" |

`ntree` |
number of trees in the random forest, see documentation from the randomForest package for details. |

`seed` |
set seed to ensure reproducibility from run to run and to standardise runs on actual and permuted data |

`...` |
Arguments to pass on to the randomForest function |

A standardised list containing

`Res.table` |
A data.frame containing significance, FDR, and the feature name. b= number of permutations yielding a higher importance than observed + 1, m= number of permutations + 1 |

`obs` |
named numeric vector, contains observed importances |

`perms` |
data.frame, contains importance values from permutations |

`Model` |
the randomForest model that was fit to the original data |

Ankur Chakravarthy

The main function is based on the idea presented in

Altmann A, Tolosi L, Sander O, Lengauer T. Permutation importance: a corrected feature importance measure. Bioinformatics. 2010 May 15;26(10):1340-7. doi: 10.1093/bioinformatics/btq134. Epub 2010 Apr 12. PubMed PMID: 20385727.

The permutation p.values in the package are exact, calculated according to

Phipson B, Smyth GK. Permutation P-values should never be zero: calculating exact P-values when permutations are randomly drawn. Stat Appl Genet Mol Biol. 2010;9:Article39. doi: 10.2202/1544-6115.1585. Epub 2010 Oct 31. PubMed PMID: 21044043.

False discovery rates account for correlations using the Two-Step BH procedure, initially reported in

Yoav Benjamini, Abba M. Krieger, and Daniel Yekutieli, 'Adaptive Linear Step-up Procedures That Control the False Discovery Rate', Biometrika, 93 (2006), 491-507.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 | ```
#Load the iris dataset
data(iris)
#Set up the predictors object
predictors=iris[,c(1:4)]
colnames(predictors)<-colnames(iris[1:4])
#Execute the main pRF function
p.test<-pRF(response=factor(iris$Species),
predictors=iris[,c(1:4)],n.perms=20,mtry=3,
type="classification",alpha=0.05)
#Put together a dataframe that consists of the
#significance stats and observed importance metrics
df<-cbind(p.test$Res.table,p.test$obs)
``` |

pRF documentation built on May 19, 2017, 3 p.m.

Questions? Problems? Suggestions? Tweet to @rdrrHQ or email at ian@mutexlabs.com.

Please suggest features or report bugs in the GitHub issue tracker.

All documentation is copyright its authors; we didn't write any of that.

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.