C-Vine Copula Structure Selection

Code Stratmill research code

Summary

This code describes a C-vine copula wrapper intended for statistical arbitrage research. It fits candidate vine structures to quantile-transformed data, restricts the candidate ordering according to a chosen target variable, and selects the structure with the lowest Akaike information criterion. An alternate ordering convention is offered because the author interprets the target variable’s position in the vine differently from the cited method.

The class also calculates conditional probabilities by numerical integration, generates simulations, and exposes AIC, BIC, and log-likelihood evaluations. The stated computational cost of enumerating candidate structures grows factorially with the number of variables, and the documentation advises keeping the universe small. This is implementation documentation rather than a trading evaluation: it supplies no return results, signal thresholds, or out-of-sample evidence, and its usefulness depends on suitable marginal transformations, data quality, and validation of the fitted dependence model.

Key ideas

Candidate C-vine orderings are fitted and ranked by AIC while respecting the selected target variable.
An alternate structure convention places the target differently in the vine ordering.
The wrapper provides conditional probability calculations, simulation, and common fit statistics.
Enumerating candidate structures has factorial cost, limiting practical use to small variable sets.
The code gives no trading performance or out-of-sample validation evidence.

Selecting C-Vine Copula Structures for Conditional Trading Analysis

Summary

Key ideas

Tags