Advanced Calibration Analysis and Tools: Identifying Influential Observations in Stochastic Interest Rate Model Calibration

Battle Count: 61

The paper is primarily relevant to quantitative risk management, actuarial science, and model validation rather than direct trading strategy development. However, it is highly relevant for: (1) calibrating interest rate models used in derivative pricing and hedging, (2) understanding parameter uncertainty in ESGs used for risk assessment, (3) identifying model risk in calibration procedures, and (4) informing calibration portfolio design. The G2++ model is widely used in Solvency II and PRIIPs contexts. The diagnostic framework could inform trading desk model validation and risk parameter estimation, but does not directly address alpha generation or trading signals.

The framework requires: (1) analytical derivation of the Jacobian and Hessian for G2++ cap pricing (complex chain rule with factorization), (2) implementation of the Weighted Hat Matrix with SVD-based pseudo-inverse, (3) Variance Stabilizing Transformations and Functional Delta Method for boundary-respecting CIs, (4) robust MAD-based scale estimation, (5) PCA for regime analysis, and (6) daily calibration pipeline over 2157 trading days. The mathematical derivations are extensive (detailed in appendix), but the factorized Jacobian implementation is computationally efficient. The R implementation is described but not publicly released.

The paper provides detailed mathematical derivations in the appendix, specifies the R implementation, uses BOBYQA solver with explicit tolerances, and describes the data processing pipeline. However, the market data from LSEG Workspace is subject to third-party licensing restrictions, and code is only available from the corresponding author on reasonable request. The analytical Jacobian factorization and all diagnostic formulas are fully specified.