Basic Usage

This section guides you through using the CopulaFurtif pipeline to create, manipulate, and diagnose bivariate copulas.

🧱 Creating Copulas

All copulas are accessible via the CopulaFactory:

from CopulaFurtif.core.copulas.domain.factories.copula_factory import CopulaFactory

copula = CopulaFactory.create("gaussian")
print(copula.name)  # Gaussian Copula

Available copulas: gaussian, student, clayton, frank, joe, gumbel, amh, tawn3, galambos, plackett, fgm, etc.

📊 Input Data

The pipeline generally expects:

• Raw data: original data for Kendall’s tau ([[X1, Y1], [X2, Y2], …])

• Pseudo-observations: data transformed to uniform scale u, v ∈ (0,1) using marginals

Generate pseudo-observations:

from CopulaFurtif.core.copulas.domain.estimation.utils import pseudo_obs

u, v = pseudo_obs(data)  # data = [[X1, Y1], [X2, Y2], ...]

📈 Accessing Basic Methods

copula.parameters = [0.5]       # or [rho, nu] for Student
print(copula.get_cdf(0.4, 0.8))
print(copula.get_pdf(0.4, 0.8))
print(copula.kendall_tau())

samples = copula.sample(100)

🔬 Diagnostics

from CopulaFurtif.core.copulas.application.services.diagnostics_service import DiagnosticService

diag = DiagnosticService()
scores = diag.evaluate(data, copula)
print(scores)

Result: a dict with LogLik, AIC, BIC, Kendall Tau Error, etc.

📌 Coming Soon: Fitting & Visualization