Control parameters for CHAID tree growing

Collects the algorithm parameters used by chaid(). The defaults match the defaults of the IBM SPSS Statistics user interface (alpha 0.05, Bonferroni adjustment on, no re-splitting, maximum depth 3, minimum parent size 100, minimum child size 50, 10 bins for continuous predictors).

Usage

chaid_control(
  alpha_merge = 0.05,
  alpha_split = 0.05,
  alpha_split_merge = 0.05,
  resplit = FALSE,
  bonferroni = TRUE,
  stat = c("pearson", "lr"),
  exhaustive_adjust = c("spss", "biggs"),
  max_depth = 3L,
  min_parent = 100,
  min_child = 50,
  min_segment = NULL,
  n_bins = 10L,
  epsilon = 0.001,
  max_iter = 100L,
  adjust_across = c("none", "bonferroni", "holm", "hochberg", "hommel", "BH", "BY")
)

Arguments

alpha_merge

Significance level for merging predictor categories. Category pairs whose test p-value exceeds this threshold are merged.

alpha_split

Significance level for splitting a node. A node is split only if the best (adjusted) p-value is at most this value.

alpha_split_merge

Significance level for re-splitting a merged compound category. Only used when resplit = TRUE.

resplit

Logical. If TRUE, compound categories consisting of three or more original categories are considered for a binary re-split during the merge step (SPSS "allow resplitting" option).

bonferroni

Logical. If TRUE (default), split p-values are Bonferroni-adjusted by the number of ways the predictor categories can be merged into the final groups.

stat

Chi-squared statistic for categorical responses: "pearson" (default) or "lr" (likelihood ratio).

exhaustive_adjust

Bonferroni multiplier convention used by Exhaustive CHAID: "spss" (default) follows the IBM SPSS algorithm document, "biggs" follows Biggs, de Ville and Suen (1991).

max_depth

Maximum tree depth (root has depth 0).

min_parent

Minimum number of cases (frequency-weighted) a node must contain to be considered for splitting.

min_child

Minimum number of cases (frequency-weighted) in each child node.

min_segment

Optional minimum size for a merged category group during the merge step (SPSS algorithm step 7). Groups smaller than this are absorbed into the most similar allowable group. NULL (default) disables this step.

n_bins

Number of quantile bins used to discretise continuous predictors.

epsilon

Convergence tolerance for the iterative estimation of expected cell frequencies with case weights.

max_iter

Maximum number of iterations for the same estimation.

adjust_across

Multiple-comparison adjustment applied across predictors within a node before comparing against alpha_split. "none" (default, as in SPSS) or one of the stats::p.adjust() methods "bonferroni", "holm", "hochberg", "hommel", "BH", "BY". Note that the adjustment family is the node, not the whole tree, and that "holm" always yields the same tree as "bonferroni" because only the minimum adjusted p-value is compared with alpha_split.

Value

An object of class "chaid_control": a list of the validated parameter values, to be passed to the control argument of chaid().

See also

chaid()

Examples

ctl <- chaid_control(max_depth = 2, min_parent = 20, min_child = 5)
fit <- chaid(Species ~ ., data = iris, control = ctl)
fit
#> CHAID decision tree (method = "chaid")
#> Response: Species (categorical) 
#> Valid cases: 150 (data: 150 rows) 
#> 
#> [1] root: setosa (33.3%), n=150 | split: Petal.Length (adj.p=1.35e-44, chi2=243.8, B=126)
#>   [2] Petal.Length in {<= 1.6}: setosa (100.0%), n=44 *
#>   [3] Petal.Length in {(1.6, 3.8]}: versicolor (57.1%), n=14 *
#>   [4] Petal.Length in {(3.8, 4.6]}: versicolor (96.9%), n=32 *
#>   [5] Petal.Length in {(4.6, 4.9]}: versicolor (64.3%), n=14 *
#>   [6] Petal.Length in {(4.9, 5.3]}: virginica (87.5%), n=16 *
#>   [7] Petal.Length in {> 5.3}: virginica (100.0%), n=30 *