You are reading a research paper published on EntryLib as separate claims, each with its evidence. How to read this page

Version 1 · 16 November 2026 · Finance

Multi-step estimators and shrinkage effect in time series models

Responsible researcher: Ivan Svetunkov (0000-0001-7826-0281)

Authors of the original paper: PersonIvan Svetunkov (0000-0001-7826-0281), PersonNikolaos Kourentzes (0000-0003-0211-5218), PersonRebecca Killick

CC-BY-4.0 · Keywords: multi-step estimators, parameter shrinkage, time series forecasting, state space models, exponential smoothing, ARIMA, predictive likelihood

Contributions
  • Ivan Svetunkov (0000-0001-7826-0281): Conceptualization, Formal analysis, Methodology, Software, Writing – original draft, Writing – review & editing
  • Gemini · claimset preparation (AI agent): Writing – original draft

Converted by its authors from their paper: The published article, Computational Statistics (2024) 39:1203-1239..

Text file for AI

Text file: the whole paper as plain text, to attach when your assistant cannot open the address.

The message

Read https://entrylib.org/p/[email protected], a research paper written as claims with their evidence. Give me a brief summary of its main findings and how well each is supported. If you cannot open the link, tell me instead of guessing. Then wait for my questions.

Summary

Forecasting models are sometimes estimated by minimising errors several steps ahead rather than one step ahead, which makes forecasts more robust when the model is misspecified. Why this happens had lacked a general explanation. Using a linear state space framework covering ARIMA and exponential smoothing models, this paper shows that multi-step loss functions automatically shrink dynamic model parameters toward zero in finite samples, whereas static regression parameters experience no shrinkage. Because strong shrinkage over long horizons can introduce estimation bias, the paper derives a General Predictive Likelihood, which avoids shrinkage when the model is correctly specified. It then proposes the Geometric Trace Mean Squared Error (GTMSE) estimator to moderate shrinkage. Simulations and real-data tests on Greenland surface temperatures and airline passengers show that GTMSE improves parameter efficiency over existing multi-step estimators and outperforms one-step estimation under model misspecification.

Highlights

  • SupportS 65Minimising sample multi-step forecast error loss functions in multi-step estimators shrinks the dynamic parameters of univariate SSOE state space models toward zero in finite samples relative to MSE_1, with shrinkage intensity increasing with the forecast horizon hh and decreasing with the sample size TT. C1
  • SupportS 61Maximising the General Predictive Likelihood yields parameter estimates identical to those of the conventional one-step-ahead estimator MSE_1 under correct model specification. C3
  • DefinitionDGTMSE: Geometric Trace Mean Squared Error estimator introduced in this paper, minimising the sum of logarithmic multi-step mean squared errors across horizons 11 to hh: GTMSE=∑j=1hlog⁡(1T−h∑t=1T−het+j|t2)GTMSE = \sum_{j=1}^{h} \log \left( \frac{1}{T-h} \sum_{t=1}^{T-h} e_{t+j|t}^2 \right) D6
  • SupportS 71The proposed GTMSE estimator imposes a milder parameter shrinkage effect than existing multi-step estimators in finite samples. C4
  • DefinitionDGeneral Predictive Likelihood: General Predictive Likelihood for consecutive observations Yt=(yt+1,…,yt+h)′Y_t = (y_{t+1}, \dots, y_{t+h})' under Gaussian errors, whose concentrated log-likelihood is lGPLh(θ,Σ^∣Y)=−T−h2(hlog⁡(2πe)+log⁡|Σ^|)l_{GPL_h}(\theta, \hat{\Sigma} \mid Y) = -\frac{T-h}{2} \left( h \log(2\pi e) + \log|\hat{\Sigma}| \right) where Σ^=1T−h∑t=1T−hEtEt′\hat{\Sigma} = \frac{1}{T-h} \sum_{t=1}^{T-h} E_t E_t' is the sample multi-step forecast error covariance matrix. D7

Context

Addresses TIAO and XU (1993), Section 1, p. 623 · from C1
Tiao and Xu (1993) observed in simulations that multi-step estimation produces more conservative models and pushes MA parameters toward boundary values on finite samples, without providing a general theoretical explanation.
Extends doi:10.1016/j.ijforecast.2004.06.002, Section 2, p. 203 · from C2
Chevillon and Hendry (2005) analysed direct multi-step estimation in dynamic autoregressions, whereas static linear regression lacks dynamic state transitions.
Extends Clements and Hendry (1998), p. 77 · from C3
Clements and Hendry (1998) formulate the multi-step predictive log-likelihood and note diagonal covariance under exact assumptions.
Addresses Stern and Coe (1984), p. 484 · from C4
Weiss and Andersen (1984) introduced the additive Trace Mean Squared Error (TMSE) multi-step loss function across horizons 1 to h.
Addresses TIAO and XU (1993), Section 3, pp. 628-633 · from C5
Tiao and Xu (1993) demonstrated via simulation that multi-step estimators lose finite-sample parameter efficiency relative to one-step MSE as horizon increases.
Extends Hoogerheide and van Dijk (2010), Section 1, p. 266 · from C6
Proietti (2011) showed that multi-step estimation gains over one-step estimation depend on model misspecification and data characteristics.
Contradicts Martinez (2017), p. 11 · from C7
Martinez (2017) assumes zero correlation between multi-step forecast errors when assumptions hold.
Extends Kourentzes N (2018) · from C9
Kourentzes and Trapero (2018) examined MSE_h and TMSE parameter effects on selected ETS models.

Definitions

Methods 14

D2 MSE_1: Conventional one-step-ahead mean squared error estimator minimising MSE1=1T−1∑t=1T−1et+1|t2MSE_1 = \frac{1}{T-1} \sum_{t=1}^{T-1} e_{t+1|t}^2 where et+1|t=yt+1−μ^t+1|te_{t+1|t} = y_{t+1} - \hat{\mu}_{t+1|t} is the one-step-ahead in-sample forecast error over TT observations.

D3 MSE_h: Multi-step mean squared error estimator minimising the sample squared error at a specific forecast horizon hh: MSEh=1T−h∑t=1T−het+h|t2MSE_h = \frac{1}{T-h} \sum_{t=1}^{T-h} e_{t+h|t}^2 where et+h|t=yt+h−μ^t+h|te_{t+h|t} = y_{t+h} - \hat{\mu}_{t+h|t} is the conditional hh-step-ahead forecast error. Source: TIAO and XU (1993).

D4 TMSE: Trace Mean Squared Error estimator minimising the sum of squared forecast errors across horizons 11 to hh: TMSE=1T−h∑j=1h∑t=1T−het+j|t2TMSE = \frac{1}{T-h} \sum_{j=1}^{h} \sum_{t=1}^{T-h} e_{t+j|t}^2 Source: Stern and Coe (1984).

D5 MSCE: Mean Squared Cumulative Error estimator minimising the squared sum of forecast errors over horizons 11 to hh: MSCE=1T−h∑t=1T−h(∑j=1het+j|t)2MSCE = \frac{1}{T-h} \sum_{t=1}^{T-h} \left( \sum_{j=1}^{h} e_{t+j|t} \right)^2 Source: doi:10.1016/j.omega.2021.102614.

D6 GTMSE: Geometric Trace Mean Squared Error estimator introduced in this paper, minimising the sum of logarithmic multi-step mean squared errors across horizons 11 to hh: GTMSE=∑j=1hlog⁡(1T−h∑t=1T−het+j|t2)GTMSE = \sum_{j=1}^{h} \log \left( \frac{1}{T-h} \sum_{t=1}^{T-h} e_{t+j|t}^2 \right)

D7 General Predictive Likelihood: General Predictive Likelihood for consecutive observations Yt=(yt+1,…,yt+h)′Y_t = (y_{t+1}, \dots, y_{t+h})' under Gaussian errors, whose concentrated log-likelihood is lGPLh(θ,Σ^∣Y)=−T−h2(hlog⁡(2πe)+log⁡|Σ^|)l_{GPL_h}(\theta, \hat{\Sigma} \mid Y) = -\frac{T-h}{2} \left( h \log(2\pi e) + \log|\hat{\Sigma}| \right) where Σ^=1T−h∑t=1T−hEtEt′\hat{\Sigma} = \frac{1}{T-h} \sum_{t=1}^{T-h} E_t E_t' is the sample multi-step forecast error covariance matrix.

D8 Generalised Variance: Determinant of the sample multi-step forecast error covariance matrix, GV=|Σ^|GV = |\hat{\Sigma}| whose minimisation is equivalent to maximising the General Predictive Likelihood.

D9 ETS(A,N,N): Pure additive local-level exponential smoothing model in the Gardner taxonomy with level smoothing parameter α\alpha, multi-step coefficient c^j=α^\hat{c}_j = \hat{\alpha}, and hh-step conditional variance sh2=s12(1+(h−1)α^2)s_h^2 = s_1^2(1 + (h-1)\hat{\alpha}^2). Source: Gardner (2006).

D10 ARIMA(0,1,1): First-order integrated moving average model (1−B)yt=(1+θB)ϵt(1-B)y_t = (1+\theta B)\epsilon_t with moving average parameter θ\theta. Source: Box G (1976).

D11 ARIMA(1,1,1): First-order autoregressive integrated moving average model (1−B)(1−ϕ1B)yt=(1+θ1B)ϵt(1-B)(1-\phi_1 B)y_t = (1+\theta_1 B)\epsilon_t with AR parameter ϕ1\phi_1 and MA parameter θ1\theta_1. Source: Box G (1976).

D12 ETS(A,A,N): Pure additive local-trend exponential smoothing model in the Gardner taxonomy with level smoothing parameter α\alpha and trend smoothing parameter β\beta, multi-step coefficient c^j=α^+β^j\hat{c}_j = \hat{\alpha} + \hat{\beta}j, and hh-step conditional variance sh2=s12(1+∑j=1h−1(α^+β^j)2)s_h^2 = s_1^2\left(1 + \sum_{j=1}^{h-1}(\hat{\alpha} + \hat{\beta}j)^2\right). Source: Gardner (2006).

D15 compared estimators: Group of all six parameter estimators compared in the empirical evaluations: MSE_1, MSE_h, TMSE, GTMSE, MSCE, and General Predictive Likelihood. Members: D2, D3, D4, D6, D5, D7.

D19 multi-step estimators: Multi-step forecast error estimators studied in the paper: MSE_h, TMSE, MSCE, and GTMSE, minimising functions of et+j|t2e_{t+j|t}^2. Members: D3, D4, D5, D6.

D20 existing multi-step estimators: Existing multi-step forecast error estimators prior to GTMSE: MSE_h, TMSE, and MSCE, minimising functions of et+j|t2e_{t+j|t}^2. Members: D3, D4, D5.

Data 2

D13 HadCRUT4 Greenland Mean Surface Temperature series: Monthly Mean Surface Temperature time series from HadCRUT.4.5.0.0 at grid cell (322.5∘,62.5∘)(322.5^\circ, 62.5^\circ) in Greenland, January 1849 to December 2017 (2,016 observations total; first 1,956 for training and final 60 for holdout evaluation).

D14 Air Passengers series: Classical monthly airline passenger totals time series from Box and Jenkins (1976) evaluated with a misspecified pure additive ETS(A,A,A)ETS(A,A,A) model over a 12-month forecast horizon. Source: Box G (1976).

Variables and measures 4

D16 RMSE: Root Mean Squared Error over forecast horizon hh: RMSE=1h∑j=1het+j|t2RMSE = \sqrt{\frac{1}{h} \sum_{j=1}^{h} e_{t+j|t}^2}

D17 MAE: Mean Absolute Error over forecast horizon hh: MAE=1h∑j=1h|et+j|t|MAE = \frac{1}{h} \sum_{j=1}^{h} |e_{t+j|t}|

D18 ME: Mean Error measuring forecast bias over forecast horizon hh: ME=1h∑j=1het+j|tME = \frac{1}{h} \sum_{j=1}^{h} e_{t+j|t}

D21 parameter estimation bias and dispersion: Distribution of estimated model parameters across 500 Monte Carlo replications evaluated visually via boxplots for median bias and interquartile range.

Theories 1

D1 SSOE state space model: An additive-error linear state space model of the form yt=w′vt−1+ϵtvt=Fvt−1+gϵt\begin{aligned} y_t &= w' v_{t-1} + \epsilon_t \\ v_t &= F v_{t-1} + g \epsilon_t \end{aligned} where yty_t is the observed value at time tt, vtv_t is the state vector, FF is the transition matrix, gg is the persistence vector, ww is the measurement vector, and ϵt∼i.i.d. 𝒩(0,σ12)\epsilon_t \sim \text{i.i.d. } \mathcal{N}(0, \sigma_1^2), with multi-step coefficients cj=w′Fj−1gc_j = w' F^{j-1} g. Source: Snyder (1985).

Assumptions

C19 · statistical · unchecked

The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. Rested on by C1, C2, C3, C4, C7, C8, C9, C10, C11, C12, C13.

C20 · statistical · unchecked

The conditional expectation of the hh-step-ahead forecast error in population is zero, E(ηt+h|t)=0E(\eta_{t+h|t}) = 0. Rested on by C1, C8, C9, C10.

C21 · statistical · unchecked

The one-step-ahead innovation errors ϵt\epsilon_t follow a Normal distribution 𝒩(0,σ12)\mathcal{N}(0, \sigma_1^2). Rested on by C3, C11, C12, C13.

Principal claims

SupportS 65C1 · formal · novelty: new, by the authors

Minimising sample multi-step forecast error loss functions in multi-step estimators shrinks the dynamic parameters of univariate SSOE state space models toward zero in finite samples relative to MSE_1, with shrinkage intensity increasing with the forecast horizon hh and decreasing with the sample size TT.

SupportS 71C2 · formal · novelty: new, by the authors

Minimising multi-step forecast error loss functions in multi-step estimators imposes no parameter shrinkage on static linear regression models.

SupportS 61C3 · formal · novelty: new, by the authors

Maximising the General Predictive Likelihood yields parameter estimates identical to those of the conventional one-step-ahead estimator MSE_1 under correct model specification.

SupportS 71C4 · formal · novelty: new, by the authors

The proposed GTMSE estimator imposes a milder parameter shrinkage effect than existing multi-step estimators in finite samples.

SupportS 39C5 · descriptive · novelty: new, by the authors

In Monte Carlo simulations of ETS(A,N,N) and ARIMA(0,1,1) processes, GTMSE produces less biased and more efficient parameter estimates with lower parameter estimation bias and dispersion than existing multi-step estimators.

SupportS 84C6 · descriptive · novelty: new, by the authors

On the HadCRUT4 Greenland Mean Surface Temperature series and the misspecified Air Passengers series, GTMSE achieves lower out-of-sample RMSE, MAE, and absolute ME than MSE_1.

Supporting claims

SupportS 72C7 · formal · novelty: new, by the authors

In an SSOE state space model with independent homoscedastic errors, the covariance σi,j\sigma_{i,j} between ii-step-ahead and jj-step-ahead forecast errors is given by σi,j={σ12c|i−j|when i=1 or j=1σ12(c|i−j|+∑l=1i−1c|l−j|c|i−l|)otherwise\sigma_{i,j} = \begin{cases} \sigma_1^2 c_{|i-j|} & \text{when } i=1 \text{ or } j=1 \\ \sigma_1^2 \left(c_{|i-j|} + \sum_{l=1}^{i-1} c_{|l-j|} c_{|i-l|}\right) & \text{otherwise} \end{cases} so multi-step forecast errors have non-zero cross-horizon covariance whenever cj≠0c_j \ne 0.

SupportS 65C8 · formal · novelty: new, by the authors

Minimising multi-step forecast error loss functions in multi-step estimators for an ARIMA(1,1,1) model shrinks both ϕ^1\hat{\phi}_1 and θ^1\hat{\theta}_1 toward −1-1 as the forecast horizon hh increases, degenerating the model into a restricted ARIMA(1,0,0)ARIMA(1,0,0) model.

SupportS 65C9 · formal · novelty: new, by the authors

Minimising multi-step loss functions in multi-step estimators shrinks smoothing parameters toward zero in pure additive exponential smoothing models, turning ETS(A,N,N), ETS(A,A,N), ETS(A,Ad,N)ETS(A,A_d,N), and ETS(A,N,A)ETS(A,N,A) into deterministic models as h→∞h \to \infty.

SupportS 65C10 · formal · novelty: new, by the authors

In ETS(A,N,A)ETS(A,N,A) with periodicity mm, minimising multi-step loss functions in multi-step estimators shrinks the seasonal smoothing parameter γ^\hat{\gamma} in a stepwise manner only when h>mh > m, and faster than α^\hat{\alpha} once h>2mh > 2m.

SupportS 71C11 · formal · novelty: new, by the authors

Under Gaussian errors, MSE_h can achieve lower asymptotic parameter variance than MSE_1 in parameter regions where cjc_j is close to zero, despite having higher variance in general.

SupportS 71C12 · formal · novelty: new, by the authors

Under Gaussian errors, the maximum likelihood estimator of MSCE is less efficient in finite samples than that of MSE_h when c^j>0\hat{c}_j > 0 for all j=1,…,h−1j = 1, \dots, h-1.

SupportS 71C13 · formal · novelty: new, by the authors

The logarithm of the Generalised Variance decomposes as log⁡GV=log⁡s12+log⁡|A^|\log GV = \log s_1^2 + \log |\hat{A}|, where |A^|=1|\hat{A}| = 1 under correct model specification.

SupportS 44C14 · descriptive · novelty: new, by the authors

In Monte Carlo simulations of ETS(A,N,N) and ARIMA(0,1,1) processes, MSE_1 produces more efficient and less biased parameter estimates with lower parameter estimation bias and dispersion than multi-step estimators when true model parameters are non-zero.

SupportS 44C15 · descriptive · novelty: new, by the authors

When fitting an over-parameterised ARIMA(1,1,1) model to ETS(A,N,N) data on small samples, multi-step estimators force the autoregressive parameter ϕ^1\hat{\phi}_1 toward negative values and produce worse parameter estimates than MSE_1.

SupportS 89C16 · descriptive · novelty: new, by the authors

On the HadCRUT4 Greenland Mean Surface Temperature series, GTMSE, TMSE, and MSCE achieve lower out-of-sample RMSE, MAE, and absolute ME than MSE_1 and General Predictive Likelihood.

SupportS 89C17 · descriptive · novelty: new, by the authors

On the misspecified Air Passengers series, GTMSE achieves the lowest out-of-sample RMSE, MAE, and ME among all compared estimators.

SupportS 99C18 · descriptive · novelty: new, by the authors

On the HadCRUT4 Greenland Mean Surface Temperature series, MSE_h with h=60h=60 achieves the lowest out-of-sample RMSE, MAE, and absolute ME among all compared estimators while shrinking the AR(1)AR(1) parameter of ARIMA(1,1,2)ARIMA(1,1,2) to −0.786-0.786.

Evidence

E1 · formal · cited by C1, C2, C7, C8, C9, C10

Analytical derivation in the single source of error (SSOE) state space framework showing that multi-step loss functions minimise multi-step forecast error variances and covariances, shrinking dynamic state-space parameters toward zero (and regression parameters not at all because g=0g=0), along with specific derivations for ARIMA(1,1,1) and pure additive ETS models (Sections 2, 3, and Appendices A–B).

  1. \eta_{t+h|t} = y_{t+h} - \mu_{t+h|t} = \sum_{j=1}^{h-1} c_j \epsilon_{t+j} + \epsilon_{t+h}, \quad c_j = w' F^{j-1} g (definition; from D1)
  2. \sigma_h^2 = \sigma_1^2 \left(1 + \sum_{j=1}^{h-1} c_j^2\right), \quad \sigma_{i,j} = \sigma_1^2 \left(c_{|i-j|} + \sum_{l=1}^{i-1} c_{|l-j|} c_{|i-l|}\right) \text{ for } i,j > 1 (variance and covariance of linear combinations; from S1, C19)
  3. \text{Minimising } MSE_h, TMSE, \text{ or } MSCE \text{ minimises } s_1^2 \text{ and } \hat{c}_j^2 \text{ toward } 0 \text{ for } h > 1, \text{ whereas } g=0 \text{ in static regression implies } c_j = 0 \text{ and no shrinkage.} (sample substitution and minimisation; from S2)

E2 · formal · cited by C3, C4, C11, C12, C13

Analytical derivation of concentrated predictive log-likelihoods for MSE_h, MSCE, and the General Predictive Likelihood (GPL), proving that log GV = log s_1^2 + log |\hat{A}| and that |\hat{A}| = 1 under correct specification, and defining GTMSE via diagonal restriction on \hat{\Sigma} (Section 4 and Appendices C–E).

  1. l_{GPL_h}(\theta, \hat{\Sigma} \mid Y) = -\frac{T-h}{2}\left(h \log(2\pi e) + \log|\hat{\Sigma}|\right), \quad GV = |\hat{\Sigma}| (concentrating multivariate normal log-likelihood; from D7, C21)
  2. \log GV = \log s_1^2 + \log|\hat{A}|, \quad |\hat{A}| = 1 \text{ under correct model specification} (matrix determinant reduction; from S1)
  3. GTMSE \propto \sum_{j=1}^{h} \log(s_j^2) = h \log(s_1^2) + \sum_{j=2}^{h} \log\left(1 + \sum_{i=1}^{j-1} \hat{c}_i^2\right) (diagonal restriction and logarithmic scaling; from S1, D6)

E3 · simulation · cited by C5, C14, C15

Monte Carlo simulation experiment generating 500 series of length 5,000 each from ETS(A,N,N) (α=0.2\alpha = 0.2) and ARIMA(0,1,1) (θ=0.6\theta = 0.6), evaluated on sub-samples T∈{20,50,100,200,500,1000,5000}T \in \{20, 50, 100, 200, 500, 1000, 5000\} and horizons h∈{10,20,50,100,200}h \in \{10, 20, 50, 100, 200\} across four fitted models (ETS(A,N,N), ETS(A,A,N), ARIMA(0,1,1), ARIMA(1,1,1)) and five estimators (MSE_1, MSE_h, TMSE, GTMSE, MSCE) using R package smooth v2.5.3 (Section 5, Figures 1–12).

E4 · quant · benchmark · secondary · n 2016 · cited by C6, C16, C18

Empirical out-of-sample evaluation of ARIMA(1,1,2) estimated with MSE_1, MSE_h (h=60), TMSE, GTMSE, MSCE, and GPL on the HadCRUT.4.5.0.0 monthly Mean Surface Temperature series for Greenland (322.5°, 62.5°), trained on 1,956 observations (Jan 1849–Dec 2012) and tested on 60 holdout observations (Jan 2013–Dec 2017) using smooth v3.2.1 in R (Section 6.1, Table 2).

E4.T1 · Out-of-sample error measures and estimated ARIMA(1,1,2) parameters on the HadCRUT4 Greenland Mean Surface Temperature series (Table 2). (HadCRUT4 Greenland Mean Surface Temperature series: 60-step holdout (Jan 2013-Dec 2017))
Rowcompared estimatorsRMSEMAEMEAR(1) parameterMA(1) parameterMA(2) parameter
r1MSE_11.2170.849-0.450.558-1.2350.265
r2MSE_h1.1280.7840.009-0.786-0.13-0.858
r3TMSE1.1980.827-0.402-0.002-0.68-0.291
r4GTMSE1.190.819-0.3660.691-1.2910.305
r5MSCE1.1850.817-0.3430.634-1.0590.077
r6General Predictive Likelihood1.2140.846-0.4440.532-1.2190.244

E5 · quant · benchmark · secondary · n 144 · cited by C6, C17

Empirical out-of-sample evaluation of a misspecified pure additive ETS(A,A,A) model estimated with MSE_1, MSE_h (h=12), TMSE, GTMSE, MSCE, and GPL on the monthly Air Passengers series over a 12-month forecast horizon using es() in smooth v3.2.1 in R (Section 6.2, Table 3).

E5.T1 · Out-of-sample error measures and estimated ETS(A,A,A) smoothing parameters on the Air Passengers series (Table 3). (Air Passengers series: 12-step holdout)
Rowcompared estimatorsRMSEMAEMElevel smoothing parameter alphatrend smoothing parameter betaseasonal smoothing parameter gamma
r1MSE_149.42337.11528.369100
r2MSE_h77.19457.98354.6190.80600.089
r3TMSE19.26615.6169.7390.12600.798
r4GTMSE18.67215.0058.3690.14500.826
r5MSCE22.60118.73613.9220.10200.546
r6General Predictive Likelihood20.01416.42611.1280.11100.846

Limitations

L1 · model specification · could bias magnitude · affects C1, C8, C9, C10 · mitigation: none

In finite samples with large forecast horizons hh, multi-step estimators can over-shrink parameters toward boundary values, introducing finite-sample estimation bias and slow convergence rates.

L2 · model specification · reduces transferability · affects C2 · mitigation: none

The zero-shrinkage analytical result applies to static linear regression models without dynamic state transitions (g=0g = 0) and does not transfer to dynamic regression models with autoregressive or moving-average error structures.

L3 · model specification · could reverse · affects C3 · mitigation: none

The equivalence between General Predictive Likelihood and one-step-ahead MSE_1 relies on correct model specification so that |A^|=1|\hat{A}| = 1; under model misspecification, off-diagonal covariances do not cancel and GPL estimates diverge from MSE_1.

L4 · model specification · reduces transferability · affects C4 · mitigation: none

The analytical derivation of GTMSE imposes a diagonal restriction setting off-diagonal forecast error covariances to zero, which does not hold in real dynamic time series processes whenever cj≠0c_j \ne 0.

L5 · sample size · reduces transferability · affects C5, C14, C15 · mitigation: none

Monte Carlo simulation results are restricted to two univariate data-generating processes (ETS(A,N,N) with α=0.2\alpha=0.2 and ARIMA(0,1,1) with θ=0.6\theta=0.6) and pure additive linear models, and numerical outcomes are reported visually via boxplots rather than tabulated summary statistics.

L6 · sampling · reduces transferability · affects C6, C16, C17, C18 · mitigation: none

Empirical evaluations are limited to two individual univariate time series (Greenland Mean Surface Temperature and Air Passengers) with single fixed holdout origins rather than a large multi-series benchmark.

Open questions

O1 · from C1, C9

How the multi-step shrinkage mechanism extends to pure multiplicative state space exponential smoothing models remains only partially explored and is not analytically established across all multiplicative forms.

References

  1. Box G, Jenkins G (1976) Time series analysis: forecasting and control. Holden-day, Oakland Used by: D10, D11, D14
  2. Clements, M., Hendry, D., 1998. Forecasting Economic Time Series. Cambridge University Press. https://doi.org/10.1017/CBO9780511599286 Used by: Context
  3. Gardner, E. S., 2006. Exponential smoothing: The state of the art—Part II. International Journal of Forecasting 22, 637–666. https://doi.org/10.1016/j.ijforecast.2006.03.005 Used by: D9, D12
  4. Hoogerheide, L., van Dijk, H. K., 2010. Bayesian forecasting of Value at Risk and Expected Shortfall using adaptive importance sampling. International Journal of Forecasting 26, 231–247. https://doi.org/10.1016/j.ijforecast.2010.01.007 Used by: Context
  5. Kourentzes N, Trapero JR (2018) On the use of multi-step cost functions for generating forecasts. Department of Management Science Working Paper Series Used by: Context
  6. Martinez, A. B., 2017. Testing for Differences in Path Forecast Accuracy: Forecast-Error Dynamics Matter. Working paper (Federal Reserve Bank of Cleveland). https://doi.org/10.26509/frbc-wp-201717 Used by: Context
  7. Snyder, R. D., 1985. Recursive Estimation of Dynamic Linear Models. Journal of the Royal Statistical Society Series B: Statistical Methodology 47, 272–276. https://doi.org/10.1111/j.2517-6161.1985.tb01355.x Used by: D1
  8. Stern, R. D., Coe, R., 1984. A Model Fitting Analysis of Daily Rainfall Data. Journal of the Royal Statistical Society. Series A (General) 147, 1. https://doi.org/10.2307/2981736 Used by: Context, D4
  9. TIAO, G. C., XU, D., 1993. Robustness of maximum likelihood estimates for multi-step predictions: The exponential smoothing case. Biometrika 80, 623–641. https://doi.org/10.1093/biomet/80.3.623 Used by: Context, D3
  10. https://doi.org/10.1016/j.ijforecast.2004.06.002 Used by: Context
  11. https://doi.org/10.1016/j.omega.2021.102614 Used by: D5

Artefacts

  • A1 · manuscript The published article, Computational Statistics (2024) 39:1203-1239. s00180-023-01377-x

Score

Support of the paper: 65, Partly supported

Scores are computed by rules from the claimset file (EntryLib rubric 0.7): they check the record, not the science, and may change as the rubric and reviews develop. The rubric. Every issue as text, for an agent preparing the next version.

The mean of its 6 principal claims; the lowest is C5 at 39. Open problems in the principal claims and what they rest on, each counted once, with the most it costs a claim:

  1. −45 Checkable Verifiability: unverifiable. In C5, C14, C15. To remove it: Make the data, code and models the claim rests on available, or state why they cannot be.
  2. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. In C1, C2, C3, C4, C5, C6, C7, C8, C9, C10, C11, C12, C13, C14, C15, C16, C17. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  3. −10 Open problems Limitation L3 (could reverse, not mitigated): The equivalence between General Predictive Likelihood and one-step-ahead MSE_1 relies on correct model specification so that |A^|=1|\hat{A}| = 1; under model misspecification, off-diagonal covariances do not cancel and GPL estimates diverge from MSE_1. In C3. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  4. −5 Plan A principal claim reported without a plan recorded beforehand. In C5, C6. To remove it: Plan the measurement in the lab record or a registration before running it, or publish the lab record that shows every run.
  5. −6 Open problems Limitation L1 (could bias magnitude, not mitigated): In finite samples with large forecast horizons hh, multi-step estimators can over-shrink parameters toward boundary values, introducing finite-sample estimation bias and slow convergence rates. In C1, C8, C9, C10. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  6. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. In C1, C2, C3, C4, C7, C8, C9, C10, C11, C12, C13. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  7. −1 Open problems Limitation L2 (reduces transferability, not mitigated): The zero-shrinkage analytical result applies to static linear regression models without dynamic state transitions (g=0g = 0) and does not transfer to dynamic regression models with autoregressive or moving-average error structures. In C2. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  8. −1 Open problems Limitation L4 (reduces transferability, not mitigated): The analytical derivation of GTMSE imposes a diagonal restriction setting off-diagonal forecast error covariances to zero, which does not hold in real dynamic time series processes whenever cj≠0c_j \ne 0. In C4. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  9. −1 Open problems Limitation L5 (reduces transferability, not mitigated): Monte Carlo simulation results are restricted to two univariate data-generating processes (ETS(A,N,N) with α=0.2\alpha=0.2 and ARIMA(0,1,1) with θ=0.6\theta=0.6) and pure additive linear models, and numerical outcomes are reported visually via boxplots rather than tabulated summary statistics. In C5, C14, C15. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  10. −1 Open problems Limitation L6 (reduces transferability, not mitigated): Empirical evaluations are limited to two individual univariate time series (Greenland Mean Surface Temperature and Air Passengers) with single fixed holdout origins rather than a large multi-series benchmark. In C6, C16, C17, C18. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  11. −1 Open problems Assumption C20 is not checked: The conditional expectation of the hh-step-ahead forecast error in population is zero, E(ηt+h|t)=0E(\eta_{t+h|t}) = 0. In C1, C8, C9, C10. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  12. −1 Open problems Assumption C21 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t follow a Normal distribution 𝒩(0,σ12)\mathcal{N}(0, \sigma_1^2). In C3, C11, C12, C13. To remove it: Test the assumption, or analyse how sensitive the claim is to it.

Claims

C1: 65, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −6 Open problems Limitation L1 (could bias magnitude, not mitigated): In finite samples with large forecast horizons hh, multi-step estimators can over-shrink parameters toward boundary values, introducing finite-sample estimation bias and slow convergence rates. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  3. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  4. −1 Open problems Assumption C20 is not checked: The conditional expectation of the hh-step-ahead forecast error in population is zero, E(ηt+h|t)=0E(\eta_{t+h|t}) = 0. To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C2: 71, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  3. −1 Open problems Limitation L2 (reduces transferability, not mitigated): The zero-shrinkage analytical result applies to static linear regression models without dynamic state transitions (g=0g = 0) and does not transfer to dynamic regression models with autoregressive or moving-average error structures. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

C3: 61, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −10 Open problems Limitation L3 (could reverse, not mitigated): The equivalence between General Predictive Likelihood and one-step-ahead MSE_1 relies on correct model specification so that |A^|=1|\hat{A}| = 1; under model misspecification, off-diagonal covariances do not cancel and GPL estimates diverge from MSE_1. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  3. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  4. −1 Open problems Assumption C21 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t follow a Normal distribution 𝒩(0,σ12)\mathcal{N}(0, \sigma_1^2). To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C4: 71, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  3. −1 Open problems Limitation L4 (reduces transferability, not mitigated): The analytical derivation of GTMSE imposes a diagonal restriction setting off-diagonal forecast error covariances to zero, which does not hold in real dynamic time series processes whenever cj≠0c_j \ne 0. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

C5: 39, Weakly supported

Novelty: new, by the authors

  1. −45 Checkable Verifiability: unverifiable. To remove it: Make the data, code and models the claim rests on available, or state why they cannot be.
  2. −10 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  3. −5 Plan A principal claim reported without a plan recorded beforehand. To remove it: Plan the measurement in the lab record or a registration before running it, or publish the lab record that shows every run.
  4. −1 Open problems Limitation L5 (reduces transferability, not mitigated): Monte Carlo simulation results are restricted to two univariate data-generating processes (ETS(A,N,N) with α=0.2\alpha=0.2 and ARIMA(0,1,1) with θ=0.6\theta=0.6) and pure additive linear models, and numerical outcomes are reported visually via boxplots rather than tabulated summary statistics. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

C6: 84, Well supported

Novelty: new, by the authors

  1. −10 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −5 Plan A principal claim reported without a plan recorded beforehand. To remove it: Plan the measurement in the lab record or a registration before running it, or publish the lab record that shows every run.
  3. −1 Open problems Limitation L6 (reduces transferability, not mitigated): Empirical evaluations are limited to two individual univariate time series (Greenland Mean Surface Temperature and Air Passengers) with single fixed holdout origins rather than a large multi-series benchmark. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

C7: 72, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C8: 65, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −6 Open problems Limitation L1 (could bias magnitude, not mitigated): In finite samples with large forecast horizons hh, multi-step estimators can over-shrink parameters toward boundary values, introducing finite-sample estimation bias and slow convergence rates. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  3. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  4. −1 Open problems Assumption C20 is not checked: The conditional expectation of the hh-step-ahead forecast error in population is zero, E(ηt+h|t)=0E(\eta_{t+h|t}) = 0. To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C9: 65, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −6 Open problems Limitation L1 (could bias magnitude, not mitigated): In finite samples with large forecast horizons hh, multi-step estimators can over-shrink parameters toward boundary values, introducing finite-sample estimation bias and slow convergence rates. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  3. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  4. −1 Open problems Assumption C20 is not checked: The conditional expectation of the hh-step-ahead forecast error in population is zero, E(ηt+h|t)=0E(\eta_{t+h|t}) = 0. To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C10: 65, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −6 Open problems Limitation L1 (could bias magnitude, not mitigated): In finite samples with large forecast horizons hh, multi-step estimators can over-shrink parameters toward boundary values, introducing finite-sample estimation bias and slow convergence rates. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.
  3. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  4. −1 Open problems Assumption C20 is not checked: The conditional expectation of the hh-step-ahead forecast error in population is zero, E(ηt+h|t)=0E(\eta_{t+h|t}) = 0. To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C11: 71, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  3. −1 Open problems Assumption C21 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t follow a Normal distribution 𝒩(0,σ12)\mathcal{N}(0, \sigma_1^2). To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C12: 71, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  3. −1 Open problems Assumption C21 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t follow a Normal distribution 𝒩(0,σ12)\mathcal{N}(0, \sigma_1^2). To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C13: 71, Partly supported

Novelty: new, by the authors

  1. −25 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −3 Open problems Assumption C19 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t of the state space model are independent and identically distributed with zero mean and constant variance σ12\sigma_1^2. To remove it: Test the assumption, or analyse how sensitive the claim is to it.
  3. −1 Open problems Assumption C21 is not checked: The one-step-ahead innovation errors ϵt\epsilon_t follow a Normal distribution 𝒩(0,σ12)\mathcal{N}(0, \sigma_1^2). To remove it: Test the assumption, or analyse how sensitive the claim is to it.

C14: 44, Weakly supported

Novelty: new, by the authors

  1. −45 Checkable Verifiability: unverifiable. To remove it: Make the data, code and models the claim rests on available, or state why they cannot be.
  2. −10 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  3. −1 Open problems Limitation L5 (reduces transferability, not mitigated): Monte Carlo simulation results are restricted to two univariate data-generating processes (ETS(A,N,N) with α=0.2\alpha=0.2 and ARIMA(0,1,1) with θ=0.6\theta=0.6) and pure additive linear models, and numerical outcomes are reported visually via boxplots rather than tabulated summary statistics. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

C15: 44, Weakly supported

Novelty: new, by the authors

  1. −45 Checkable Verifiability: unverifiable. To remove it: Make the data, code and models the claim rests on available, or state why they cannot be.
  2. −10 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  3. −1 Open problems Limitation L5 (reduces transferability, not mitigated): Monte Carlo simulation results are restricted to two univariate data-generating processes (ETS(A,N,N) with α=0.2\alpha=0.2 and ARIMA(0,1,1) with θ=0.6\theta=0.6) and pure additive linear models, and numerical outcomes are reported visually via boxplots rather than tabulated summary statistics. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

C16: 89, Well supported

Novelty: new, by the authors

  1. −10 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −1 Open problems Limitation L6 (reduces transferability, not mitigated): Empirical evaluations are limited to two individual univariate time series (Greenland Mean Surface Temperature and Air Passengers) with single fixed holdout origins rather than a large multi-series benchmark. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

C17: 89, Well supported

Novelty: new, by the authors

  1. −10 Evidence Rests on an unchecked proof; evidence named without the observations the claim reports. To remove it: Support the claim with evidence the claimset records, citing the result entries or table rows it reports, or have its proof checked; for cited work, read the work in full and quote the passage the claim rests on from a captured copy (claimset extension review).
  2. −1 Open problems Limitation L6 (reduces transferability, not mitigated): Empirical evaluations are limited to two individual univariate time series (Greenland Mean Surface Temperature and Air Passengers) with single fixed holdout origins rather than a large multi-series benchmark. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

C18: 99, Well supported

Novelty: new, by the authors

  1. −1 Open problems Limitation L6 (reduces transferability, not mitigated): Empirical evaluations are limited to two individual univariate time series (Greenland Mean Surface Temperature and Air Passengers) with single fixed holdout origins rather than a large multi-series benchmark. To remove it: Mitigate it with evidence, or narrow the claim's scope so that it no longer applies.

Versions

  • Version 1 (this page) · 16 November 2026 at 20:13 UTC · active

Every version is kept in the public registry (github.com/entrylib/registry).