anofox-regression 0.5.13

A robust statistics library for regression analysis
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
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254

#!/usr/bin/env Rscript
# Generate validation data for ALM implementation
# This script creates test cases with known outputs from R's greybox package

library(greybox)

# Set seed for reproducibility
set.seed(42)

# Helper function to print results in Rust-compatible format
print_results <- function(name, model, x, y) {
  cat(sprintf("\n// %s\n", name))
  cat(sprintf("// Distribution: %s\n", model$distribution))

  # Print data
  cat("// X data:\n")
  cat(sprintf("// %s\n", paste(round(x, 6), collapse=", ")))
  cat("// Y data:\n")
  cat(sprintf("// %s\n", paste(round(y, 6), collapse=", ")))

  # Coefficients
  coefs <- coef(model)
  cat(sprintf("// Intercept: %.10f\n", coefs[1]))
  if(length(coefs) > 1) {
    cat(sprintf("// Coefficients: %s\n", paste(sprintf("%.10f", coefs[-1]), collapse=", ")))
  }

  # Scale parameter
  cat(sprintf("// Scale (sigma): %.10f\n", model$scale))

  # Log-likelihood
  cat(sprintf("// Log-likelihood: %.10f\n", logLik(model)))

  # AIC/BIC
  cat(sprintf("// AIC: %.10f\n", AIC(model)))
  cat(sprintf("// BIC: %.10f\n", BIC(model)))

  # Fitted values (first 5)
  fitted_vals <- fitted(model)
  cat(sprintf("// Fitted values (first 5): %s\n",
              paste(sprintf("%.10f", head(fitted_vals, 5)), collapse=", ")))

  # Residuals (first 5)
  resids <- residuals(model)
  cat(sprintf("// Residuals (first 5): %s\n",
              paste(sprintf("%.10f", head(resids, 5)), collapse=", ")))

  cat("\n")
}

cat("// =============================================================================\n")
cat("// ALM Validation Data Generated from R greybox package\n")
cat("// Generated on:", format(Sys.time()), "\n")
cat("// greybox version:", as.character(packageVersion("greybox")), "\n")
cat("// =============================================================================\n\n")

# -----------------------------------------------------------------------------
# Test Case 1: Normal distribution (simple linear regression)
# -----------------------------------------------------------------------------
n <- 50
x1 <- seq(1, 50, length.out = n)
y1 <- 2.5 + 1.5 * x1 + rnorm(n, sd = 3)

model_normal <- alm(y1 ~ x1, distribution = "dnorm")
print_results("Normal Distribution", model_normal, x1, y1)

# Print Rust test case
cat("// Rust test for Normal:\n")
cat(sprintf("const X_NORMAL: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", x1), collapse=", ")))
cat(sprintf("const Y_NORMAL: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y1), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_NORMAL: f64 = %.10f;\n", coef(model_normal)[1]))
cat(sprintf("const EXPECTED_COEF_NORMAL: f64 = %.10f;\n", coef(model_normal)[2]))
cat(sprintf("const EXPECTED_SCALE_NORMAL: f64 = %.10f;\n", model_normal$scale))
cat(sprintf("const EXPECTED_LL_NORMAL: f64 = %.10f;\n", logLik(model_normal)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 2: Laplace distribution (robust regression)
# -----------------------------------------------------------------------------
y2 <- 2.5 + 1.5 * x1 + rnorm(n, sd = 3)
# Add outliers
y2[10] <- y2[10] + 50
y2[40] <- y2[40] - 50

model_laplace <- alm(y2 ~ x1, distribution = "dlaplace")
print_results("Laplace Distribution", model_laplace, x1, y2)

cat("// Rust test for Laplace:\n")
cat(sprintf("const Y_LAPLACE: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y2), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_LAPLACE: f64 = %.10f;\n", coef(model_laplace)[1]))
cat(sprintf("const EXPECTED_COEF_LAPLACE: f64 = %.10f;\n", coef(model_laplace)[2]))
cat(sprintf("const EXPECTED_SCALE_LAPLACE: f64 = %.10f;\n", model_laplace$scale))
cat(sprintf("const EXPECTED_LL_LAPLACE: f64 = %.10f;\n", logLik(model_laplace)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 3: Student-t distribution
# -----------------------------------------------------------------------------
y3 <- 2.5 + 1.5 * x1 + rt(n, df = 5) * 3

model_t <- alm(y3 ~ x1, distribution = "dt")
print_results("Student-t Distribution", model_t, x1, y3)

cat("// Rust test for Student-t:\n")
cat(sprintf("const Y_STUDENT_T: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y3), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_T: f64 = %.10f;\n", coef(model_t)[1]))
cat(sprintf("const EXPECTED_COEF_T: f64 = %.10f;\n", coef(model_t)[2]))
cat(sprintf("const EXPECTED_SCALE_T: f64 = %.10f;\n", model_t$scale))
cat(sprintf("const EXPECTED_LL_T: f64 = %.10f;\n", logLik(model_t)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 4: Log-Normal distribution
# -----------------------------------------------------------------------------
y4 <- exp(0.5 + 0.03 * x1 + rnorm(n, sd = 0.3))

model_lnorm <- alm(y4 ~ x1, distribution = "dlnorm")
print_results("Log-Normal Distribution", model_lnorm, x1, y4)

cat("// Rust test for Log-Normal:\n")
cat(sprintf("const Y_LOGNORMAL: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y4), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_LOGNORM: f64 = %.10f;\n", coef(model_lnorm)[1]))
cat(sprintf("const EXPECTED_COEF_LOGNORM: f64 = %.10f;\n", coef(model_lnorm)[2]))
cat(sprintf("const EXPECTED_SCALE_LOGNORM: f64 = %.10f;\n", model_lnorm$scale))
cat(sprintf("const EXPECTED_LL_LOGNORM: f64 = %.10f;\n", logLik(model_lnorm)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 5: Poisson distribution (count data)
# -----------------------------------------------------------------------------
x5 <- seq(0, 2, length.out = n)
lambda <- exp(0.5 + 1.0 * x5)
y5 <- rpois(n, lambda)

model_pois <- alm(y5 ~ x5, distribution = "dpois")
print_results("Poisson Distribution", model_pois, x5, y5)

cat("// Rust test for Poisson:\n")
cat(sprintf("const X_POISSON: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", x5), collapse=", ")))
cat(sprintf("const Y_POISSON: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y5), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_POISSON: f64 = %.10f;\n", coef(model_pois)[1]))
cat(sprintf("const EXPECTED_COEF_POISSON: f64 = %.10f;\n", coef(model_pois)[2]))
cat(sprintf("const EXPECTED_LL_POISSON: f64 = %.10f;\n", logLik(model_pois)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 6: Negative Binomial distribution
# -----------------------------------------------------------------------------
# Use same x5
mu_nb <- exp(0.5 + 0.8 * x5)
y6 <- rnbinom(n, size = 2, mu = mu_nb)

model_nbinom <- alm(y6 ~ x5, distribution = "dnbinom")
print_results("Negative Binomial Distribution", model_nbinom, x5, y6)

cat("// Rust test for Negative Binomial:\n")
cat(sprintf("const Y_NEGBINOM: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y6), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_NEGBINOM: f64 = %.10f;\n", coef(model_nbinom)[1]))
cat(sprintf("const EXPECTED_COEF_NEGBINOM: f64 = %.10f;\n", coef(model_nbinom)[2]))
cat(sprintf("const EXPECTED_LL_NEGBINOM: f64 = %.10f;\n", logLik(model_nbinom)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 7: Gamma distribution
# -----------------------------------------------------------------------------
shape <- 2
rate <- shape / exp(0.5 + 0.05 * x1)
y7 <- rgamma(n, shape = shape, rate = rate)

model_gamma <- alm(y7 ~ x1, distribution = "dgamma")
print_results("Gamma Distribution", model_gamma, x1, y7)

cat("// Rust test for Gamma:\n")
cat(sprintf("const Y_GAMMA: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y7), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_GAMMA: f64 = %.10f;\n", coef(model_gamma)[1]))
cat(sprintf("const EXPECTED_COEF_GAMMA: f64 = %.10f;\n", coef(model_gamma)[2]))
cat(sprintf("const EXPECTED_SCALE_GAMMA: f64 = %.10f;\n", model_gamma$scale))
cat(sprintf("const EXPECTED_LL_GAMMA: f64 = %.10f;\n", logLik(model_gamma)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 8: Inverse Gaussian distribution
# -----------------------------------------------------------------------------
mu_ig <- exp(0.5 + 0.03 * x1)
y8 <- statmod::rinvgauss(n, mean = mu_ig, shape = 2)

model_ig <- alm(y8 ~ x1, distribution = "dinvgauss")
print_results("Inverse Gaussian Distribution", model_ig, x1, y8)

cat("// Rust test for Inverse Gaussian:\n")
cat(sprintf("const Y_INVGAUSS: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y8), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_INVGAUSS: f64 = %.10f;\n", coef(model_ig)[1]))
cat(sprintf("const EXPECTED_COEF_INVGAUSS: f64 = %.10f;\n", coef(model_ig)[2]))
cat(sprintf("const EXPECTED_LL_INVGAUSS: f64 = %.10f;\n", logLik(model_ig)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 9: Asymmetric Laplace (Quantile regression at 0.75)
# -----------------------------------------------------------------------------
y9 <- 2.5 + 1.5 * x1 + rnorm(n, sd = 3)

model_alaplace <- alm(y9 ~ x1, distribution = "dalaplace", alpha = 0.75)
print_results("Asymmetric Laplace (alpha=0.75)", model_alaplace, x1, y9)

cat("// Rust test for Asymmetric Laplace (quantile = 0.75):\n")
cat(sprintf("const Y_ALAPLACE: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y9), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_ALAPLACE: f64 = %.10f;\n", coef(model_alaplace)[1]))
cat(sprintf("const EXPECTED_COEF_ALAPLACE: f64 = %.10f;\n", coef(model_alaplace)[2]))
cat(sprintf("const EXPECTED_SCALE_ALAPLACE: f64 = %.10f;\n", model_alaplace$scale))
cat(sprintf("const EXPECTED_LL_ALAPLACE: f64 = %.10f;\n", logLik(model_alaplace)))
cat("\n")

# -----------------------------------------------------------------------------
# Test Case 10: Logistic distribution
# -----------------------------------------------------------------------------
y10 <- 2.5 + 1.5 * x1 + rlogis(n, scale = 2)

model_logis <- alm(y10 ~ x1, distribution = "dlogis")
print_results("Logistic Distribution", model_logis, x1, y10)

cat("// Rust test for Logistic:\n")
cat(sprintf("const Y_LOGISTIC: [f64; %d] = [%s];\n", n, paste(sprintf("%.6f", y10), collapse=", ")))
cat(sprintf("const EXPECTED_INTERCEPT_LOGISTIC: f64 = %.10f;\n", coef(model_logis)[1]))
cat(sprintf("const EXPECTED_COEF_LOGISTIC: f64 = %.10f;\n", coef(model_logis)[2]))
cat(sprintf("const EXPECTED_SCALE_LOGISTIC: f64 = %.10f;\n", model_logis$scale))
cat(sprintf("const EXPECTED_LL_LOGISTIC: f64 = %.10f;\n", logLik(model_logis)))
cat("\n")

# -----------------------------------------------------------------------------
# Summary statistics comparison
# -----------------------------------------------------------------------------
cat("// =============================================================================\n")
cat("// Summary of expected values for validation tests\n")
cat("// =============================================================================\n")
cat(sprintf("// Normal:       intercept=%.6f, coef=%.6f, scale=%.6f, LL=%.6f\n",
            coef(model_normal)[1], coef(model_normal)[2], model_normal$scale, logLik(model_normal)))
cat(sprintf("// Laplace:      intercept=%.6f, coef=%.6f, scale=%.6f, LL=%.6f\n",
            coef(model_laplace)[1], coef(model_laplace)[2], model_laplace$scale, logLik(model_laplace)))
cat(sprintf("// Student-t:    intercept=%.6f, coef=%.6f, scale=%.6f, LL=%.6f\n",
            coef(model_t)[1], coef(model_t)[2], model_t$scale, logLik(model_t)))
cat(sprintf("// Log-Normal:   intercept=%.6f, coef=%.6f, scale=%.6f, LL=%.6f\n",
            coef(model_lnorm)[1], coef(model_lnorm)[2], model_lnorm$scale, logLik(model_lnorm)))
cat(sprintf("// Poisson:      intercept=%.6f, coef=%.6f, LL=%.6f\n",
            coef(model_pois)[1], coef(model_pois)[2], logLik(model_pois)))
cat(sprintf("// NegBinom:     intercept=%.6f, coef=%.6f, LL=%.6f\n",
            coef(model_nbinom)[1], coef(model_nbinom)[2], logLik(model_nbinom)))
cat(sprintf("// Gamma:        intercept=%.6f, coef=%.6f, scale=%.6f, LL=%.6f\n",
            coef(model_gamma)[1], coef(model_gamma)[2], model_gamma$scale, logLik(model_gamma)))
cat(sprintf("// InvGauss:     intercept=%.6f, coef=%.6f, LL=%.6f\n",
            coef(model_ig)[1], coef(model_ig)[2], logLik(model_ig)))
cat(sprintf("// AsymLaplace:  intercept=%.6f, coef=%.6f, scale=%.6f, LL=%.6f\n",
            coef(model_alaplace)[1], coef(model_alaplace)[2], model_alaplace$scale, logLik(model_alaplace)))
cat(sprintf("// Logistic:     intercept=%.6f, coef=%.6f, scale=%.6f, LL=%.6f\n",
            coef(model_logis)[1], coef(model_logis)[2], model_logis$scale, logLik(model_logis)))