SVARtca 1.0.3

No numerical results change in this release. tca_systems_form(), tca_analyze(), tca_decompose_binary(), tca_from_var() and plot_tca() return exactly the same values as in version 1.0.2. The changes concern the additivity diagnostic, documentation and attribution.

• Author attribution corrected. The methodology paper is by Enrico Wegner, Lenard Lieb, Stephan Smeekes and Ines Wilms (arXiv:2405.18987, doi:10.48550/arXiv.2405.18987). Earlier versions cited “Emanuel Wegner” and omitted Ines Wilms. Corrected in DESCRIPTION (including the contributor entries: Enrico Wegner is credited as author of the MATLAB TCA toolbox the package is ported from and as co-author of the methodology; Lieb, Smeekes and Wilms as authors of the methodology), in all help pages, the vignette and the paper drafts; the CITATION version note was updated.
• tca_validate_additivity() now performs a real check of Theorem 2(ii) of the paper. Previously it computed through = total - not_through and then tested total - (through + not_through), which is zero by construction and could not fail. The through effect is now recomputed independently with AND conditions (a first-passage partition of the paths through the variable) and through + not_through is compared with total at every response variable other than the conditioning variable. New argument pair additionally checks the inclusion-exclusion identity through(v1 and v2) = through(v1) + through(v2) - through(v1 or v2) used by the 4-way mode, with the left-hand side computed directly by AND conditions on both variables. New argument tol (default 1e-10). The function also reports whether B is strictly lower triangular. The return value is still an invisible logical, now with attributes residuals, pair_residual, max_residual and lower_triangular. On a valid systems form the residual is of the order of 1e-16; on a corrupted B (for example a backward edge above the diagonal) the residual is nonzero and the check fails.
• print.tca_result() documentation corrected: the default target is variable 1 (the first variable in the original ordering), as the code has always done, not the first intermediate.
• tca_systems_form() documentation: the Psis argument is now documented, with a note that the matrices must be the reduced-form MA coefficients on the reduced-form errors u_t, not the structural MA matrices on the structural shocks (Psi_j = Phi0 Psi^s_j Phi0^{-1}). Passing structural matrices gives wrong impulse responses at horizons 1 and above.
• tca_decompose_binary() documentation now states that total = through + not_through holds by construction and points to tca_validate_additivity() for the independent check.
• Tests: the additivity tests can now fail, and the 3-way and 4-way tests compare every channel with a brute-force enumeration of the paths in the systems form graph (K = 4, h = 2) instead of checking a sum that holds by construction; new tests check that a corrupted B is reported, that the inclusion-exclusion identity holds, that tca_decompose_binary() matches an independent AND-based through effect and a hand-computed IRF recursion, and that the worked example of Section 2.2 of the paper, evaluated at the package’s own parameter values alpha1 = -0.4, alpha2 = 0.3, alpha3 = 0.5, alpha4 = 1.5, gives TE = 0.6884057971, IE = 0.2811094453 and DE = 0.4072963518, equal to the paper’s symbolic formulas evaluated at the same values.

SVARtca 1.0.0

• Initial release (renamed from TCA to SVARtca to avoid CRAN name conflict).
• Implements Transmission Channel Analysis (Wegner, Lieb, Smeekes and Wilms 2025) [citation corrected in 1.0.3].
• Three decomposition modes: overlapping, exhaustive 3-way, exhaustive 4-way.
• Systems form construction from VAR coefficient matrices.
• Integration with the vars package via tca_from_var().
• Plotting with plot_tca() (ggplot2).
• Binary additivity validation with tca_validate_additivity().