linreg-core 0.8.1

Lightweight regression library (OLS, Ridge, Lasso, Elastic Net, WLS, LOESS, Polynomial) with 14 diagnostic tests, cross validation, and prediction intervals. Pure Rust - no external math dependencies. WASM, Python, FFI, and Excel XLL bindings.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132

# ============================================================================
# White Test Reference Implementation (R)
# ============================================================================
# This script generates reference values for the White test using R's
# skedastic::white function. The White test is a more general test for
# heteroscedasticity that uses squares and cross-products of predictors.
#
# Source: skedastic package, white function
# Reference: White (1980), "A Heteroskedasticity-Consistent Covariance
#            Matrix Estimator and a Direct Test for Heteroskedasticity"
#
# Usage:
#   Rscript test_white.R [csv_path] [output_dir]
#   Args:
#     csv_path  - Path to CSV file (first col = response, rest = predictors)
#                 Default: ../../datasets/csv/mtcars.csv
#     output_dir- Path to output directory
#                 Default: ../../results/r
# ============================================================================

library(skedastic)
library(jsonlite)

# Helper function to convert categorical columns to numeric
convert_categorical_to_numeric <- function(data, dataset_name) {
  # Check for non-numeric columns
  non_numeric_cols <- names(data)[sapply(data, function(x) !is.numeric(x))]

  if (length(non_numeric_cols) > 0) {
    cat(paste0("INFO: Dataset '", dataset_name, "' contains non-numeric columns: ",
               paste(non_numeric_cols, collapse = ", "), "\n"))
    cat("Converting categorical variables to numeric representations...\n")

    for (col in non_numeric_cols) {
      if (is.factor(data[[col]])) {
        # For factors, use integer level encoding
        unique_vals <- length(unique(data[[col]]))
        data[[col]] <- as.numeric(data[[col]])
        cat(paste0("  ", col, ": ", unique_vals, " unique values -> integer level encoding\n"))
      } else if (is.character(data[[col]])) {
        # For character columns, first try to convert to numeric directly
        # If that produces NAs, convert to factor and use integer encoding
        unique_vals <- length(unique(data[[col]]))
        temp_numeric <- as.numeric(data[[col]])
        if (any(is.na(temp_numeric))) {
          # Contains non-numeric strings, use factor encoding
          data[[col]] <- as.numeric(as.factor(data[[col]]))
          cat(paste0("  ", col, ": ", unique_vals, " unique values -> integer level encoding\n"))
        } else {
          # Successfully converted to numeric
          data[[col]] <- temp_numeric
          cat(paste0("  ", col, ": ", unique_vals, " unique values -> numeric encoding\n"))
        }
      }
    }
  }

  return(data)
}

# Parse command line arguments
args <- commandArgs(trailingOnly = TRUE)

# Set defaults
default_csv <- "../../datasets/csv/mtcars.csv"
default_output <- "../../results/r"

# Parse arguments
csv_path <- ifelse(length(args) >= 1, args[1], default_csv)
output_dir <- ifelse(length(args) >= 2, args[2], default_output)

# Validate CSV path
if (!file.exists(csv_path)) {
  stop(paste("CSV file not found:", csv_path))
}

# Extract dataset name from filename
dataset_name <- tools::file_path_sans_ext(basename(csv_path))

# Read CSV data
data <- read.csv(csv_path)

# Convert categorical columns to numeric
data <- convert_categorical_to_numeric(data, dataset_name)

# Assume first column is response variable, rest are predictors
response_col <- names(data)[1]
predictor_cols <- names(data)[-1]

# Build formula: response ~ predictor1 + predictor2 + ...
formula_str <- paste(response_col, "~", paste(predictor_cols, collapse = " + "))
formula <- as.formula(formula_str)

# Fit the model
model <- lm(formula, data = data)

# Run White test
# white() parameters:
# - mainlm: the model
# - interactions: FALSE for original variables only (default)
white_result <- white(model, interactions = FALSE)

# Print results
cat("White Test (R - skedastic::white)\n")
cat("=================================\n")
cat("Dataset:", dataset_name, "\n")
cat("Formula:", formula_str, "\n")
cat("LM-statistic:", white_result$statistic, "\n")
cat("p-value:", white_result$p.value, "\n")
cat("Passed:", white_result$p.value > 0.05, "\n\n")

# Prepare output
output <- list(
  test_name = "White Test (R - skedastic::white)",
  dataset = dataset_name,
  formula = formula_str,
  statistic = as.numeric(white_result$statistic),
  p_value = as.numeric(white_result$p.value),
  passed = white_result$p.value > 0.05,
  description = "Tests for heteroscedasticity using squares and cross-products of predictors. R variant uses skedastic::white with original variables only."
)

# Create output directory if it doesn't exist
if (!dir.exists(output_dir)) {
  dir.create(output_dir, recursive = TRUE)
}

# Save to JSON with naming convention: {dataset}_white.json
output_file <- file.path(output_dir, paste0(dataset_name, "_white.json"))
write_json(output, output_file, pretty = TRUE, digits = 22)

cat("Results saved to:", output_file, "\n")