interaction_investigator {bartMachine} | R Documentation |

Explore the pairwise interaction counts for a BART model to learn about interactions fit by the model. This function includes an option to generate a plot of the pairwise interaction counts.

interaction_investigator(bart_machine, plot = TRUE, num_replicates_for_avg = 5, num_trees_bottleneck = 20, num_var_plot = 50, cut_bottom = NULL, bottom_margin = 10)

`bart_machine` |
An object of class “bartMachine”. |

`plot` |
If TRUE, a plot of the pairwise interaction counts is generated. |

`num_replicates_for_avg` |
The number of replicates of BART to be used to generate pairwise interaction inclusion counts. Averaging across multiple BART models improves stability of the estimates. |

`num_trees_bottleneck` |
Number of trees to be used in the sum-of-trees model for computing pairwise interactions counts. A small number of trees should be used to force the variables to compete for entry into the model. |

`num_var_plot` |
Number of variables to be shown on the plot. If “Inf,” all variables are plotted (not recommended if the number of predictors is large). Default is 50. |

`cut_bottom` |
A display parameter between 0 and 1 that controls where the y-axis is plotted. A value of 0 would begin the y-axis at 0; a value of 1 begins the y-axis at the minimum of the average pairwise interaction inclusion count (the smallest bar in the bar plot). Values between 0 and 1 begin the y-axis as a percentage of that minimum. |

`bottom_margin` |
A display parameter that adjusts the bottom margin of the graph if labels are clipped. The scale of this parameter is the same as set with |

An interaction between two variables is considered to occur whenever a path from any node of a tree to any of its terminal node contains splits using those two variables. See Kapelner and Bleich, 2013, Section 4.11.

`interaction_counts_avg` |
For each of the |

`interaction_counts_sd` |
For each of the |

In the plot, the red bars correspond to the standard error of the variable inclusion proportion estimates (since multiple replicates were used).

Adam Kapelner and Justin Bleich

Adam Kapelner, Justin Bleich (2016). bartMachine: Machine Learning with Bayesian Additive Regression Trees. Journal of Statistical Software, 70(4), 1-40. doi:10.18637/jss.v070.i04

## Not run: #generate Friedman data set.seed(11) n = 200 p = 10 X = data.frame(matrix(runif(n * p), ncol = p)) y = 10 * sin(pi* X[ ,1] * X[,2]) +20 * (X[,3] -.5)^2 + 10 * X[ ,4] + 5 * X[,5] + rnorm(n) ##build BART regression model bart_machine = bartMachine(X, y, num_trees = 20) #investigate interactions interaction_investigator(bart_machine) ## End(Not run)

[Package *bartMachine* version 1.2.6 Index]