pith. sign in

arxiv: 2408.10610 · v5 · submitted 2024-08-20 · 💻 cs.LG · math.PR· stat.ME

On an L² norm for stationary ARMA processes

Anand Ganesh , Babhrubahan Bose , Anand Rajagopalan This is my paper

Pith reviewed 2026-05-23 22:13 UTC · model grok-4.3

classification 💻 cs.LG math.PRstat.ME
keywords ARMA processesL2 normWold decompositionmean square prediction errorstationary time seriesautoregressive modelsmoving average models
0
0 comments X

The pith

Stationary ARMA models admit an L² norm computed from the Wold decomposition of an underlying true process.

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

The paper aims to introduce a norm that places different stationary ARMA models into a common space so they can be compared or bounded directly. It embeds the ARMA models inside the Hilbert space generated by the past and present of a true linearly non-deterministic stationary process X_t. The norm itself is obtained by projecting onto the Wold decomposition of that true process. One concrete payoff is explicit bounds on the mean square prediction error incurred when an AR(1) model is fitted to an MA(1) process. The authors then check whether those bounds hold on generated sample paths.

Core claim

We propose an L² norm for stationary ARMA models. We look at ARMA models within the Hilbert space of the past with present of a true purely linearly non-deterministic stationary process X_t, and compute the L² norm based on its Wold decomposition. As an application of this L² norm, we derive bounds on the mean square prediction error for AR(1) models of MA(1) processes, and verify these bounds empirically for sample data.

What carries the argument

L² norm on ARMA models induced by the Wold decomposition of the true embedding process X_t

If this is right

  • Explicit upper and lower bounds follow for the mean square prediction error when an AR(1) model is used on an MA(1) process.
  • The same norm supplies a distance that lets ARMA models be compared inside one Hilbert space.
  • Empirical checks on finite samples become a direct test of the derived prediction-error bounds.

Where Pith is reading between the lines

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

  • The same construction could be used to bound prediction error for other low-order ARMA pairs without new derivations.
  • Because the norm lives in the Wold space of a single true process, it offers a route to consistency results when the true process is only approximately ARMA.
  • The empirical verification step shows the bounds remain informative even when the fitted model order is deliberately misspecified.

Load-bearing premise

ARMA models can be embedded inside the Hilbert space generated by the past and present of a true purely linearly non-deterministic stationary process X_t.

What would settle it

Sample paths from an MA(1) process where the observed mean square prediction error of the fitted AR(1) model exceeds the upper bound derived from the proposed L² norm.

Figures

Figures reproduced from arXiv: 2408.10610 by Anand Ganesh, Anand Rajagopalan, Babhrubahan Bose.

Figure 1
Figure 1. Figure 1: Compare Pad´e and non-Pad´e errors on the unit [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
read the original abstract

We propose an $L^2$ norm for stationary Autoregressive Moving Average (ARMA) models. We look at ARMA models within the Hilbert space of the past with present of a true purely linearly non-deterministic stationary process $X_t$, and compute the $L^2$ norm based on its Wold decomposition. As an application of this $L^2$ norm, we derive bounds on the mean square prediction error for AR(1) models of MA(1) processes, and verify these bounds empirically for sample data.

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.

Referee Report

2 major / 0 minor

Summary. The paper proposes an L² norm on stationary ARMA models by embedding an ARMA process inside the Hilbert space H(X) generated by the past and present of an underlying purely linearly non-deterministic stationary process X_t and then taking the L² norm induced by the Wold decomposition of X_t. As an application it derives bounds on the mean-square prediction error when an AR(1) is used to approximate an MA(1) and checks the bounds on sample data.

Significance. If the proposed norm can be shown to be independent of the auxiliary process X_t and the derived bounds are rigorous, the construction would supply a geometrically motivated distance on the ARMA parameter space together with explicit approximation-error controls; the empirical verification on sample data is a positive feature.

major comments (2)
  1. [Abstract] Abstract and title give no indication that the numerical value of the proposed L² norm is invariant under replacement of the auxiliary process X_t by any other process that admits the same ARMA coefficients as its linear predictor. The construction is defined relative to a fixed X_t; without an invariance proof the object is not an intrinsic norm on the ARMA parameter space.
  2. [Abstract] The abstract states that bounds are derived and checked empirically, yet supplies no derivation steps, no explicit error analysis, and no description of the sample-data regime. Central claim therefore cannot be verified from available text.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and the constructive comments on invariance and abstract clarity. We address each point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract and title give no indication that the numerical value of the proposed L² norm is invariant under replacement of the auxiliary process X_t by any other process that admits the same ARMA coefficients as its linear predictor. The construction is defined relative to a fixed X_t; without an invariance proof the object is not an intrinsic norm on the ARMA parameter space.

    Authors: We agree that the abstract does not mention invariance and that the construction is presented relative to a fixed X_t. The norm is intended to depend only on the ARMA coefficients via the shared linear predictor structure in the Wold decomposition, but we did not supply an explicit invariance argument. In the revision we will insert a short proposition and proof establishing independence from the choice of auxiliary X_t (provided the ARMA coefficients match), thereby confirming the object is intrinsic to the parameter space. revision: yes

  2. Referee: [Abstract] The abstract states that bounds are derived and checked empirically, yet supplies no derivation steps, no explicit error analysis, and no description of the sample-data regime. Central claim therefore cannot be verified from available text.

    Authors: The abstract is necessarily brief. The full manuscript derives the MSPE bounds in Section 3 using the Hilbert-space embedding and Wold decomposition, with the explicit error analysis following from the orthogonality of innovations. The empirical verification appears in Section 4 on simulated MA(1) series with stated parameter values and sample sizes. To address the referee’s concern we will expand the abstract with one sentence summarizing the derivation route and the simulation regime. revision: yes

Circularity Check

0 steps flagged

No circularity: norm defined via external Wold theorem on independent X_t

full rationale

The construction embeds ARMA models inside H(X) for an external true process X_t and computes the L² norm from the Wold decomposition of that X_t. Wold's theorem is a standard external result, not derived or cited from the authors' prior work. No equations, predictions, or bounds in the abstract or description reduce by construction to fitted inputs, self-definitions, or self-citation chains. The mean-square error bounds and empirical verification are downstream applications that do not feed back into the norm definition. The derivation is therefore self-contained against external mathematical benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Ledger extracted solely from the abstract; full paper may introduce additional parameters or assumptions.

axioms (1)
  • domain assumption The underlying process X_t is purely linearly non-deterministic and stationary.
    Explicitly invoked as the setting in which the Hilbert space and Wold decomposition are applied.
invented entities (1)
  • L² norm for stationary ARMA models no independent evidence
    purpose: To quantify ARMA models inside the Hilbert space of past and present observations
    Introduced by the authors as the central object of the paper.

pith-pipeline@v0.9.0 · 5622 in / 1340 out tokens · 50670 ms · 2026-05-23T22:13:28.983535+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

  1. [1]

    Time series analysis: forecasting and control

    George EP Box and Gwilym M Jenkins. Time series analysis: forecasting and control . Holden Day, 1976

    work page 1976

  2. [2]

    The Wold decomposition theo- rem, 2014

    Umberto Triacca. The Wold decomposition theo- rem, 2014. URL https://www.lem.sssup.it/phd/ documents/Lesson11.pdf. Lecture Notes

    work page 2014

  3. [3]

    Measuring the distance between sets of ARMA models

    Umberto Triacca. Measuring the distance between sets of ARMA models. Econometrics, 2016

    work page 2016

  4. [4]

    A metric for ARMA processes

    Richard J Martin. A metric for ARMA processes. IEEE transactions on Signal Processing , 2002

    work page 2002

  5. [5]

    Statistical Analysis of Stationary Time Series

    Ulf Grenander and Murray Rosenblatt. Statistical Analysis of Stationary Time Series . Wiley Publica- tions, 1957

    work page 1957

  6. [6]

    Estimation of vector ARMAX models

    Edward J Hannan, William TM Dunsmuir, and Man- fred Deistler. Estimation of vector ARMAX models. Journal of Multivariate Analysis , 1980

    work page 1980

  7. [7]

    Some properties of the parameterization of ARMA systems with unknown or- der

    M Deistler and EJ Hannan. Some properties of the parameterization of ARMA systems with unknown or- der. Journal of Multivariate Analysis , 1981

    work page 1981

  8. [8]

    Rational distributed lag functions

    Dale W Jorgenson. Rational distributed lag functions. Econometrica: Journal of the Econometric Society , 1966

    work page 1966

  9. [9]

    A distance measure for classifying ARIMA models

    Domenico Piccolo. A distance measure for classifying ARIMA models. Journal of time series analysis, 1990

    work page 1990

  10. [10]

    Szego’s theorem and its proba- bilistic descendants

    Nicholas H Bingham. Szego’s theorem and its proba- bilistic descendants. Probability Surveys, 2012

    work page 2012

  11. [11]

    AR and MA representation of partial autocorrelation functions, with applications

    Akihiko Inoue. AR and MA representation of partial autocorrelation functions, with applications. Proba- bility theory and related fields , 2008

    work page 2008

  12. [12]

    ARMA approximations and representations of a stationary time series

    Mohsen Pourahmadi. ARMA approximations and representations of a stationary time series. Sankhya: The Indian Journal of Statistics, Series B , 1992

    work page 1992

  13. [13]

    Time Series Analysis

    James Douglas Hamilton. Time Series Analysis . Princeton University Press, 1994

    work page 1994

  14. [14]

    Shumway and D.S

    R.H. Shumway and D.S. Stoffer. Time Series Analysis and Its Applications: With R Examples . Springer, 2006

    work page 2006

  15. [15]

    Shift invariant spaces and prediction theory

    Pesi Masani. Shift invariant spaces and prediction theory. Acta Mathematica, 1962

    work page 1962

  16. [16]

    A Hilbert space problem book

    Paul R Halmos. A Hilbert space problem book . Springer Verlag, 1982

    work page 1982

  17. [17]

    The prediction the- ory of multivariate stochastic processes

    Norbert Wiener and Pesi Masani. The prediction the- ory of multivariate stochastic processes. Acta mathe- matica, 98, 1957

    work page 1957

  18. [18]

    Wold decomposi- tion, prediction and parameterization of stationary processes with infinite variance.Probability theory and related fields, 1988

    AG Miamee and M Pourahmadi. Wold decomposi- tion, prediction and parameterization of stationary processes with infinite variance.Probability theory and related fields, 1988

    work page 1988

  19. [19]

    Sequence not in l1 satisfying certain bound- edness conditions, 2015

    zhw. Sequence not in l1 satisfying certain bound- edness conditions, 2015. URL https://math. stackexchange.com/a/1291988

    work page arXiv 2015

  20. [20]

    A Course in Functional Analysis

    John B Conway. A Course in Functional Analysis . Springer, 1997

    work page 1997

  21. [21]

    Linear control theory with an H∞ optimality criterion

    JC Doyle and B Francis. Linear control theory with an H∞ optimality criterion. SIAM J. Control and Optimization, 1987

    work page 1987

  22. [22]

    Linear systems

    Thomas Kailath. Linear systems. Prentice-Hall En- glewood Cliffs, NJ, 1980. 11

    work page 1980