Influence Function: Local Robustness and Efficiency

Ruonan Xu; Xiye Yang

arxiv: 2501.15307 · v2 · submitted 2025-01-25 · 💰 econ.EM

Influence Function: Local Robustness and Efficiency

Xiye Yang , Ruonan Xu This is my paper

Pith reviewed 2026-05-23 04:47 UTC · model grok-4.3

classification 💰 econ.EM

keywords influence functionsRiesz representerorthogonal projectionmean-zero functionslocal robustnesssemiparametric estimationplug-in estimatorsefficiency equivalence

0 comments

The pith

The influence function is a linear transformation of the identification function after orthogonal projection onto mean-zero functions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper presents a differentiation-based framework for deriving influence functions that applies uniformly to parametric, nonparametric, and semiparametric models. The key step is to obtain the Riesz representer by projecting the identification function orthogonally onto the mean-zero function subspace. The influence function then follows as a linear transformation of this centered function. The approach handles infinite-dimensional parameters without change in form and, for multi-step plug-in estimators, produces an explicit adjustment that ensures local robustness. It further identifies conditions under which joint estimation matches the efficiency of plug-in methods even with over-identified nuisances.

Core claim

The Riesz representer of the functional derivative is obtained by orthogonally projecting the identification function onto the subspace of mean-zero functions. Consequently, the influence function emerges as a linear transformation of this centered moment function. The approach extends seamlessly to infinite-dimensional parameters, revealing a common algebraic form for influence functions across both finite- and infinite-dimensional parameters. Applied to semiparametric multi-step plug-in estimation, the method automatically yields locally robust moment functions and provides an explicit closed-form expression for the adjustment term. The framework also establishes verifiable sufficient cond

What carries the argument

Orthogonal projection of the identification function onto the mean-zero subspace, serving as the Riesz representer of the functional derivative from which the influence function is derived via linear transformation.

If this is right

Influence functions share the same algebraic form for finite- and infinite-dimensional parameters.
Multi-step plug-in estimators receive an explicit closed-form adjustment that produces locally robust moment functions.
Verifiable sufficient conditions are supplied for semiparametric efficiency equivalence between joint estimation and plug-in estimation when nuisances are over-identified.
The direct differentiation approach unifies derivation of influence functions across parametric, nonparametric, and semiparametric models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The projection step could be implemented numerically to approximate influence functions in high-dimensional or complex models.
The explicit adjustment term might guide construction of new estimators that enforce local robustness by design.
The framework's emphasis on verifiable conditions could be tested in applied settings with over-identified nuisances to compare efficiency in practice.

Load-bearing premise

The functional derivative exists in a suitable space and the orthogonal projection onto the mean-zero subspace is well-defined and yields the Riesz representer for both finite- and infinite-dimensional parameters.

What would settle it

A concrete model in which the functional derivative exists but the orthogonal projection of the identification function fails to recover the known influence function would disprove the unification claim.

read the original abstract

This paper introduces a direct differentiation-based framework that unifies the derivation of influence functions across parametric, nonparametric, and semiparametric models. We show that the Riesz representer of the functional derivative is obtained by orthogonally projecting the identification function onto the subspace of mean-zero functions. Consequently, the influence function emerges as a linear transformation of this centered moment function. The approach extends seamlessly to infinite-dimensional parameters, revealing a common algebraic form for influence functions across both finite- and infinite-dimensional parameters. Applied to semiparametric multi-step plug-in estimation, our method automatically yields locally robust moment functions and provides an explicit closed-form expression for the adjustment term. Finally, we leverage this framework to revisit the joint versus plug-in estimation debate, establishing verifiable sufficient conditions for their semiparametric efficiency equivalence even when nuisance parameters are over-identified.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This is a clean but largely standard re-derivation of influence functions via differentiation and projection onto mean-zero functions, with useful applications to multi-step estimators.

read the letter

The paper's main move is to derive the influence function by taking the functional derivative of the identification function and then orthogonally projecting it onto the mean-zero subspace to get the Riesz representer. From there the influence function is just a linear transformation of that centered object. They show this works for both finite- and infinite-dimensional parameters and then apply it to semiparametric multi-step plug-in problems to produce locally robust moments plus an explicit adjustment term. Finally they give verifiable conditions under which joint estimation and plug-in estimation achieve the same semiparametric efficiency bound even with over-identified nuisances. That last part is the most concrete payoff. The algebraic form they emphasize is already familiar from the classic semiparametric literature, so the unification claim does not feel like a large departure. The extension to infinite dimensions is asserted to be seamless, but the paper would need to spell out the precise function spaces and regularity conditions that make the projection well-defined in those cases. The multi-step application and the joint-versus-plug-in conditions look like the parts that could be cited by people working on robust estimation routines. The derivations appear internally consistent on the terms given, with no obvious circularity or invented objects. This is the kind of paper that belongs in an econometrics journal that publishes methodological notes on efficiency bounds. A reader who already knows the Riesz-representer route will not learn much new machinery, but someone implementing multi-step estimators might pick up the closed-form adjustment and the efficiency-equivalence conditions. It is worth sending to referees who can check the infinite-dimensional details and the sharpness of the new conditions.

Referee Report

2 major / 2 minor

Summary. The paper introduces a differentiation-based framework for deriving influence functions that unifies parametric, nonparametric, and semiparametric models. It claims that the Riesz representer of the functional derivative is obtained via orthogonal projection of the identification function onto the mean-zero subspace, yielding the influence function as a linear transformation of the centered moment function. The approach is extended to infinite-dimensional parameters and applied to multi-step plug-in estimation to produce locally robust moments with an explicit closed-form adjustment; it also derives sufficient conditions for semiparametric efficiency equivalence between joint and plug-in estimation even under over-identification of nuisances.

Significance. If the central projection construction and its extension to infinite dimensions are rigorously established with appropriate regularity conditions, the framework could streamline the derivation of influence functions and robust moment conditions in semiparametric econometrics, offering an algebraic unification across model classes and explicit adjustment terms that may facilitate efficiency comparisons. The explicit closed-form adjustment for plug-in estimators is a potential strength if the derivations hold without circularity.

major comments (2)

[Abstract / Introduction] The abstract states that the Riesz representer is obtained by orthogonally projecting the identification function onto the mean-zero subspace, but the manuscript must explicitly define the functional derivative, the relevant function space, and verify that the projection is well-defined and yields the representer for both finite- and infinite-dimensional cases (as required for the unification claim). Without these details in a dedicated section with equations, the extension to infinite dimensions remains unverified.
[Application to plug-in estimation] § on semiparametric multi-step plug-in estimation: the claim of an 'explicit closed-form expression for the adjustment term' that automatically yields locally robust moment functions requires a concrete derivation showing how the projection produces the adjustment without relying on already-known influence functions; this step is load-bearing for the applied contribution.

minor comments (2)

Notation for the identification function and the linear transformation should be introduced with explicit definitions early in the paper to improve readability.
The discussion of joint versus plug-in estimation would benefit from a brief comparison table or example contrasting the derived conditions with existing results in the literature.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major comment below and indicate planned revisions to strengthen the manuscript.

read point-by-point responses

Referee: [Abstract / Introduction] The abstract states that the Riesz representer is obtained by orthogonally projecting the identification function onto the mean-zero subspace, but the manuscript must explicitly define the functional derivative, the relevant function space, and verify that the projection is well-defined and yields the representer for both finite- and infinite-dimensional cases (as required for the unification claim). Without these details in a dedicated section with equations, the extension to infinite dimensions remains unverified.

Authors: We agree that a dedicated section with explicit definitions and verification would strengthen the presentation of the unification claim. In the revised manuscript we will add a new section that defines the functional derivative, specifies the relevant function space (including the mean-zero subspace), states the required regularity conditions, and verifies that the orthogonal projection yields the Riesz representer for both finite- and infinite-dimensional parameters, with supporting equations. revision: yes
Referee: [Application to plug-in estimation] § on semiparametric multi-step plug-in estimation: the claim of an 'explicit closed-form expression for the adjustment term' that automatically yields locally robust moment functions requires a concrete derivation showing how the projection produces the adjustment without relying on already-known influence functions; this step is load-bearing for the applied contribution.

Authors: The existing derivation obtains the adjustment term directly from the orthogonal projection of the identification function onto the mean-zero subspace. To address the concern, we will expand the section with an explicit step-by-step algebraic derivation that starts from the identification function and arrives at the closed-form adjustment without presupposing the influence function, thereby clarifying the independence of the argument. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The paper presents a differentiation-plus-orthogonal-projection construction for influence functions that follows directly from standard Riesz representation in the relevant function space. The abstract and strongest claim describe an algebraic identity (projection of the identification function onto mean-zero functions yields the representer) that is not shown to reduce to any fitted parameter, self-citation chain, or renamed input. No equations or sections in the provided material exhibit a load-bearing step that is equivalent to its own inputs by construction. The framework is therefore treated as self-contained against external benchmarks in semiparametric theory.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available, so no specific free parameters, axioms, or invented entities can be identified from the provided information.

pith-pipeline@v0.9.0 · 5662 in / 1150 out tokens · 27556 ms · 2026-05-23T04:47:08.197517+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean absolute_floor_iff_bare_distinguishability unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

the Riesz representer of the functional derivative is obtained by orthogonally projecting the identification function onto the subspace of mean-zero functions
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

influence function emerges as a linear transformation of this centered moment function

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.