detectXOR (version 0.1.0)

detect_xor: Detect XOR Patterns in Variable Pairs

Description

Identifies XOR-shaped relationships between variables using statistical tests and pattern detection.

Usage

detect_xor(
  data,
  class_col = "class",
  check_tau = TRUE,
  compute_axes_parallel_significance = TRUE,
  p_threshold = 0.05,
  tau_threshold = 0.3,
  abs_diff_threshold = 20,
  split_method = "quantile",
  max_cores = 1,
  extreme_handling = "winsorize",
  winsor_limits = c(0.05, 0.95),
  scale_data = TRUE,
  use_complete = TRUE
)

Value

List containing:

results_df

Data frame with detection results for all variable pairs

pair_list

Detailed analysis results for each variable pair

Arguments

data

Data frame containing features and class column

class_col

Name of class column (default: "class")

check_tau

Logical - compute classwise tau coefficients (default: TRUE)

compute_axes_parallel_significance

Logical - compute Wilcoxon tests (default: TRUE)

p_threshold

Significance threshold (default: 0.05)

tau_threshold

Tau coefficient threshold (default: 0.3)

abs_diff_threshold

Absolute difference threshold for patterns (default: 20)

split_method

Method for splitting data ("quantile" or "range") (default: "quantile")

max_cores

Maximum cores for parallel processing (default: NULL = automatic)

extreme_handling

Method for handling extreme values; options include "winsorize" or "none" (default: "winsorize")

winsor_limits

Numeric vector of length 2 specifying lower and upper quantiles for winsorization (default: c(0.05, 0.95))

scale_data

Logical; whether to scale/standardize the data before analysis (default: TRUE)

use_complete

Logical; whether to use only complete cases (default: TRUE)

Details

This function performs an analysis to detect XOR-like patterns in pairwise variable relationships within two-class data sets. The analysis pipeline includes:

1. Data preprocessing (winsorization, scaling, complete cases)

2. Tile pattern analysis using chi-squared tests

3. Classwise Kendall tau correlation analysis

4. Group-wise Wilcoxon significance tests

The function automatically handles parallel processing when multiple cores are available and returns both a summary data frame and detailed results for further analysis.

See Also

generate_spaghetti_plot_from_results for spaghetti plot visualization, generate_xy_plot_from_results for scatter plot visualization, generate_xor_reportConsole for console reporting, generate_xor_reportHTML for HTML report generation, XOR_data for example dataset

Examples

# \donttest{
# Load example data
data(XOR_data)

# Run XOR detection
results <- detect_xor(data = XOR_data, class_col = "class")

# View summary of detected patterns
print(results$results_df["xor_shape_detected"])

# Generate visualizations
spaghetti_plot <- generate_spaghetti_plot_from_results(
  results = results,
  data = XOR_data,
  class_col = "class"
)

print(spaghetti_plot)

xy_plot <- generate_xy_plot_from_results(
  results = results,
  data = XOR_data,
  class_col = "class"
)

print(xy_plot)

# Generate console report (doesn't write files)
generate_xor_reportConsole(results, XOR_data, "class", show_plots = FALSE)

# View detailed results for detected pairs
detected_pairs <- results$results_df[results$results_df$xor_shape_detected == TRUE, ]
print(detected_pairs)

# }