Regression

R-style Formula

from vectrix import regress

model = regress(data=df, formula="sales ~ ads + price + promo")
print(model.summary())

Methods

model = regress(data=df, formula="sales ~ ads + price", method="ridge", alpha=1.0)

Full Parameters

model = regress(
    y=None,              # ndarray (direct mode)
    X=None,              # ndarray (direct mode)
    data=None,           # DataFrame (formula mode)
    formula=None,        # "y ~ x1 + x2"
    method='ols',        # 'ols', 'ridge', 'lasso', 'huber', 'quantile'
    summary=True,        # auto-print summary
    alpha=None,          # regularization strength
    diagnostics=False    # auto-run diagnostics
)

Results

camelCase is the primary naming. snake_case aliases are available for backward compatibility.

print(model.rSquared)         # R² (primary)
print(model.adjRSquared)      # Adjusted R² (primary)
print(model.fStat)            # F-statistic (primary)
print(model.durbinWatson)     # Durbin-Watson (primary)
print(model.coefficients)     # Coefficient array
print(model.pvalues)          # P-values array

# snake_case aliases also work
print(model.r_squared)
print(model.adj_r_squared)
print(model.f_stat)
print(model.durbin_watson)

Diagnostics

print(model.diagnose())

Returns a text report with

• VIF: Multicollinearity check (>10 is problematic)
• Breusch-Pagan: Heteroscedasticity test
• Jarque-Bera: Residual normality test
• Durbin-Watson: Autocorrelation test

Prediction

import pandas as pd

new_data = pd.DataFrame({
    "ads": [50, 75, 90],
    "price": [20, 15, 10],
    "promo": [0, 1, 1],
})
predictions = model.predict(new_data)

Formula Syntax

regress(data=df, formula="y ~ x1 + x2")       # Specific variables
regress(data=df, formula="y ~ .")              # All variables
regress(data=df, formula="y ~ x1 * x2")       # Interaction terms
regress(data=df, formula="y ~ x + I(x**2)")   # Polynomial

Direct Array Input

model = regress(y=y_array, X=X_array)