A frequency domain approach for local module identification in dynamic networks

Karthik R. Ramaswamy, Péter Zoltán Csurcsia, Johan Schoukens, Paul M.J. Van den Hof

incompletemedium confidence

Category: Not specified
Journal tier: Specialist/Solid
Processed: Sep 28, 2025, 12:56 AM
arXiv Links: Abstract ↗PDF ↗

Audit review

The paper proposes a practical two-step frequency-domain method (nonparametric LPM-based FRFs followed by parametric weighted LS), explains Property 1 for predictor selection, and reports simulation-based benefits, but it does not supply formal proofs of pointwise consistency or variance optimality; the candidate model provides a correct proof sketch for these claims under standard assumptions (excitation rank, independence, and correct/consistent weights), filling the theoretical gaps left by the paper.

Referee report (LaTeX)

\textbf{Recommendation:} major revisions

\textbf{Journal Tier:} specialist/solid

\textbf{Justification:}

The paper makes a practical and potentially impactful contribution by combining nonparametric FRF estimation with a parametric, variance-weighted fit to identify a local module while avoiding nuisance-module parametrization. The exposition is clear, and simulations are persuasive. However, the core claims of (i) pointwise consistency of the two-stage estimator and (ii) variance benefits of weighting are not supported by formal theory. Including concise, assumption-explicit propositions with proofs would significantly strengthen the paper.