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
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238

# ============================================================================
# Extended Datasets Validation - R Reference Implementation
# ============================================================================
# This script validates the Rust OLS implementation against R using synthetic
# and real-world datasets designed to test edge cases and stress conditions.
#
# Output: verification/datasets/results/r/*.json
# ============================================================================

cat("==============================================================\n")
cat("Extended Datasets Validation - R Reference Implementation\n")
cat("==============================================================\n")

# Add user library to path (helps find packages installed via RGUI)
user_lib <- file.path(Sys.getenv("USERPROFILE"), "Documents", "R", "win-library", "4.4")
if (dir.exists(user_lib)) {
    .libPaths(c(user_lib, .libPaths()))
}

# Load required packages
required_packages <- c("lmtest", "car", "tseries")
for (pkg in required_packages) {
    if (!require(pkg, character.only = TRUE, quietly = TRUE)) {
        stop(sprintf("Package '%s' is required. Install with: install.packages('%s')", pkg, pkg))
    }
}

# Set up output directory (relative to script location)
script_dir <- dirname(sys.frame(1)$filename)
output_dir <- file.path(script_dir, "..", "..", "..", "results", "r")
dir.create(output_dir, recursive = TRUE, showWarnings = FALSE)

# Set path to datasets (relative to script location)
datasets_dir <- file.path(script_dir, "..", "..", "..", "datasets", "csv")

# ============================================================================
# Helper Functions
# ============================================================================

format_vector <- function(vec) {
    if (is.null(vec) || length(vec) == 0) return("[]")
    return(sprintf("[%s]", paste(vec, collapse = ", ")))
}

run_regression <- function(dataset_name, df, y_var, x_vars = NULL) {
    cat(sprintf("\n=== %s ===\n", dataset_name))

    # Build formula
    if (is.null(x_vars)) {
        x_vars <- setdiff(names(df), y_var)
    }
    formula_str <- paste(y_var, "~", paste(x_vars, collapse = " + "))
    formula <- as.formula(formula_str)

    # Fit model (with error handling)
    model <- tryCatch(lm(formula, data = df), error = function(e) {
        cat(sprintf("  Error: %s\n", conditionMessage(e)))
        return(NULL)
    })

    if (is.null(model)) return(NULL)

    # Extract results
    n <- nrow(df)
    k <- length(x_vars)
    df_residual <- model$df.residual

    coefficients <- as.vector(coef(model))
    std_errors <- as.vector(summary(model)$coefficients[, "Std. Error"])
    t_stats <- as.vector(summary(model)$coefficients[, "t value"])
    p_values <- as.vector(summary(model)$coefficients[, "Pr(>|t|)"])

    r_squared <- summary(model)$r.squared
    adj_r_squared <- summary(model)$adj.r.squared

    # Handle cases where F-statistic might be NA (e.g., perfect fit)
    f_stat <- tryCatch(summary(model)$fstatistic, error = function(e) NULL)
    if (!is.null(f_stat)) {
        f_statistic <- f_stat[1]
        f_p_value <- pf(f_statistic[1], f_statistic[2], f_statistic[3], lower.tail = FALSE)
    } else {
        f_statistic <- NA
        f_p_value <- NA
    }

    # Confidence intervals
    ci <- confint(model, level = 0.95)
    ci_lower <- as.vector(ci[, 1])
    ci_upper <- as.vector(ci[, 2])

    # VIF for multiple regression
    vif_results <- NULL
    if (k >= 2) {
        vif_values <- tryCatch(car::vif(model), error = function(e) NULL)
        if (!is.null(vif_values)) {
            vif_results <- list(
                variables = names(vif_values),
                vif = as.numeric(vif_values)
            )
        }
    }

    # Output JSON
    json_output <- sprintf('{
  "dataset": "%s",
  "n": %d,
  "k": %d,
  "df_residual": %d,
  "coefficients": %s,
  "std_errors": %s,
  "t_stats": %s,
  "p_values": %s,
  "r_squared": %.15f,
  "adj_r_squared": %.15f,
  "f_statistic": %s,
  "f_p_value": %s,
  "ci_lower": %s,
  "ci_upper": %s,
  "vif": %s
}',
        dataset_name,
        n, k, df_residual,
        format_vector(coefficients),
        format_vector(std_errors),
        format_vector(t_stats),
        format_vector(p_values),
        r_squared, adj_r_squared,
        ifelse(is.na(f_statistic), "null", as.character(f_statistic)),
        ifelse(is.na(f_p_value), "null", as.character(f_p_value)),
        format_vector(ci_lower),
        format_vector(ci_upper),
        if (!is.null(vif_results)) {
            sprintf('{"variables": ["%s"], "vif": %s}',
                    paste(vif_results$variables, collapse = '", "'),
                    paste(vif_results$vif, collapse = ", "))
        } else {
            "null"
        }
    )

    # Write to file
    output_file <- file.path(output_dir, paste0(gsub(" ", "_", tolower(dataset_name)), ".json"))
    writeLines(json_output, output_file)
    cat(sprintf("  Wrote: %s\n", basename(output_file)))

    # Print summary
    cat(sprintf("    n = %d, k = %d\n", n, k))
    cat(sprintf("    R² = %.6f, Adj R² = %.6f\n", r_squared, adj_r_squared))
    cat(sprintf("    F(%d, %d) = %.4f, p = %.6f\n", k, df_residual, f_statistic, f_p_value))
    if (!is.null(vif_results)) {
        cat(sprintf("    VIF: %s\n", paste(round(vif_results$vif, 2), collapse = ", ")))
    }

    invisible(list(
        dataset = dataset_name,
        n = n,
        k = k,
        r_squared = r_squared,
        vif = if (!is.null(vif_results)) vif_results$vif else NULL
    ))
}

# ============================================================================
# Load and Test Datasets
# ============================================================================

# --- Synthetic Datasets ---

# 1. Simple Linear
synthetic_simple <- read.csv(file.path(datasets_dir, "synthetic_simple_linear.csv"))
run_regression("Synthetic Simple Linear", synthetic_simple, "y", c("x"))

# 2. Multiple Regression
synthetic_multiple <- read.csv(file.path(datasets_dir, "synthetic_multiple.csv"))
run_regression("Synthetic Multiple", synthetic_multiple, "y", c("x1", "x2", "x3"))

# 3. Collinear (should fail or show high VIF)
synthetic_collinear <- read.csv(file.path(datasets_dir, "synthetic_collinear.csv"))
cat("\n--- Testing Collinear Dataset (expecting singular matrix or extreme VIF) ---\n")
tryCatch({
    run_regression("Synthetic Collinear", synthetic_collinear, "y", c("x1", "x2", "x3"))
}, error = function(e) {
    cat(sprintf("  Expected error: %s\n", conditionMessage(e)))
})

# 4. Heteroscedastic
synthetic_hetero <- read.csv(file.path(datasets_dir, "synthetic_heteroscedastic.csv"))
run_regression("Synthetic Heteroscedastic", synthetic_hetero, "y", c("x"))

# 5. Nonlinear
synthetic_nonlinear <- read.csv(file.path(datasets_dir, "synthetic_nonlinear.csv"))
run_regression("Synthetic Nonlinear", synthetic_nonlinear, "y", c("x"))

# 6. Nonnormal
synthetic_nonnormal <- read.csv(file.path(datasets_dir, "synthetic_nonnormal.csv"))
run_regression("Synthetic Nonnormal", synthetic_nonnormal, "y", c("x"))

# 7. Autocorrelated
synthetic_auto <- read.csv(file.path(datasets_dir, "synthetic_autocorrelated.csv"))
run_regression("Synthetic Autocorrelated", synthetic_auto, "y", c("x"))

# 8. High VIF
synthetic_high_vif <- read.csv(file.path(datasets_dir, "synthetic_high_vif.csv"))
run_regression("Synthetic High VIF", synthetic_high_vif, "y", NULL)

# 9. Outliers
synthetic_outliers <- read.csv(file.path(datasets_dir, "synthetic_outliers.csv"))
run_regression("Synthetic Outliers", synthetic_outliers, "y", NULL)

# 10. Small Sample
synthetic_small <- read.csv(file.path(datasets_dir, "synthetic_small.csv"))
run_regression("Synthetic Small", synthetic_small, "y", NULL)

# --- Real-World Datasets ---

# Longley (famous multicollinearity test case)
longley <- read.csv(file.path(datasets_dir, "longley.csv"))
# Use Employed as dependent, other columns as predictors
longley_y <- "Employed"
longley_x <- setdiff(names(longley), c(longley_y, "Year", "Unnamed..0"))
run_regression("Longley", longley, longley_y, longley_x)

# Mtcars
mtcars <- read.csv(file.path(datasets_dir, "mtcars.csv"))
run_regression("Mtcars", mtcars, "mpg", c("cyl", "disp", "hp", "wt", "qsec"))

# Bodyfat
bodyfat <- read.csv(file.path(datasets_dir, "bodyfat.csv"))
run_regression("Bodyfat", bodyfat, names(bodyfat)[1], NULL)

# Prostate
prostate <- read.csv(file.path(datasets_dir, "prostate.csv"))
run_regression("Prostate", prostate, names(prostate)[ncol(prostate)], NULL)

cat("\n==============================================================\n")
cat("Extended validation complete!\n")
cat(sprintf("Results saved to: %s\n", output_dir))
cat("==============================================================\n")