pith. sign in

arxiv: 2604.09259 · v1 · submitted 2026-04-10 · 📊 stat.ME · stat.AP

Exact Bayesian Planning for Simple Step-Stress Accelerated Life Testing with Competing Risks

Kiran Prajapat This is my paper

Pith reviewed 2026-05-10 17:01 UTC · model grok-4.3

classification 📊 stat.ME stat.AP
keywords Bayesian optimal designstep-stress accelerated life testingcompeting risksWeibull distributionpreposterior variancequantile estimationaccelerated life testing
0
0 comments X

The pith

Bayesian planning for step-stress tests with competing risks minimizes the preposterior variance of the use-stress lifetime quantile without asymptotic approximations.

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

This paper develops a Bayesian approach to design simple step-stress accelerated life tests for items that can fail from two independent competing modes. It reparametrizes the Weibull model parameters into quantile-related quantities that engineers can interpret and assign priors to directly. The optimal test plan is chosen by minimizing the exact preposterior variance of the p-th quantile of lifetime at normal use stress, using Monte Carlo simulation over possible designs and full posterior sampling. This exact method avoids large-sample approximations and remains reliable even with small numbers of test units. The approach is shown on data from a solar lighting device, where designs favoring lower stress levels near use conditions emerge consistently.

Core claim

The authors establish a planning procedure for step-stress accelerated life testing under competing risks by reparametrizing each failure mode's Weibull parameters to the p-th quantile at use stress and the scale parameter, enabling prior elicitation from engineering knowledge. They then locate the design that minimizes the preposterior variance of that quantile, evaluated by integrating over the posterior distribution using the No-U-Turn Sampler without relying on asymptotic normality. This yields optimal stress levels and change times that are valid for any sample size, as demonstrated through application to a real dataset and extensive sensitivity analyses.

What carries the argument

Quantile-based reparametrization of competing Weibull risks under the cumulative exposure model, which separates parameters into interpretable quantities for prior specification and allows exact computation of preposterior quantile variance.

If this is right

  • The optimal lower stress level consistently operates close to use conditions, independent of prior choices.
  • The optimal time to change stress shows moderate sensitivity to the target quantile probability and sample size.
  • Posterior inference remains feasible via standard MCMC even for small samples.
  • The framework extends quantile reparametrization from single failure mode to competing risks settings.
  • Sensitivity analysis confirms robustness to hyperparameter choices in the priors.

Where Pith is reading between the lines

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

  • Extending this to more than two failure modes would require generalizing the reparametrization while keeping priors elicitable.
  • Replacing the grid search for designs with a continuous optimization routine could improve computational efficiency for larger design spaces.
  • The method could inform sequential testing strategies where designs adapt based on early failure data.
  • Validation on simulated data where true parameters are known would test whether the chosen designs indeed reduce quantile uncertainty more than alternatives.

Load-bearing premise

The two competing failure modes are independent, each following a Weibull distribution with a log-linear relationship between stress and life under the cumulative exposure model.

What would settle it

Running the recommended optimal design on a new set of units and finding that the resulting posterior variance of the use-stress quantile exceeds the variance obtained from a different design point in the candidate grid.

Figures

Figures reproduced from arXiv: 2604.09259 by Kiran Prajapat.

Figure 1
Figure 1. Figure 1: EDF-based goodness-of-fit diagnostics: empirical and model-based fitted CDFs (left panel) and [PITH_FULL_IMAGE:figures/full_fig_p017_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Histograms of the MLEs for the reparametrised parameters [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Convergence diagnosis: Trace, density, and ACF plots. [PITH_FULL_IMAGE:figures/full_fig_p022_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: SA1: Smoothed criterion functions C1(τ ) and C2(τ ) under varying quantile probability p ∈ {0.01, 0.10, 0.50}. Vertical dashed lines indicate the optimal τ 0 for each criterion. 6.1.2 SA2: Sensitivity with Respect to Lower Stress Level s1 The lower stress level s1 determines the thermal conditions experienced by units during the first phase of the step-stress experiment. In practice, s1 is a design choice … view at source ↗
Figure 5
Figure 5. Figure 5: SA2: Smoothed criterion functions C1(τ ) and C2(τ ) for five representative lower stress levels s1 ∈ {1/298, 1/309, 1/320, 1/333, 1/346} K −1 . Vertical red dashed lines indicate the optimal τ 0 for each criterion. 27 [PITH_FULL_IMAGE:figures/full_fig_p027_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: SA3: Smoothed criterion functions C1(τ ) and C2(τ ) under Prior choices; Prior I, Prior II and Prior III. Vertical dashed lines in red indicate the optimal τ 0 for each criterion. 29 [PITH_FULL_IMAGE:figures/full_fig_p029_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: SA4: Smoothed criterion functions C1(τ ) and C2(τ ) under varying sample size n ∈ {20, 35, 50}. Vertical dashed lines in red indicate the optimal τ 0 for each criterion. 6.2 Two-Variable Optimal Design The two-variable optimal design is obtained by minimising C1(D) and C2(D) jointly over both s1 and τ using Algorithm 2, with p = 0.10, n = 35, and B = 1000. The results are reported in [PITH_FULL_IMAGE:figu… view at source ↗
Figure 8
Figure 8. Figure 8: Smoothed criterion surfaces C1(D) and C2(D) over the two-dimensional design space (x1, τ ) under Prior I (tighter, top) and Prior II (wider, bottom). Each surface is shown from two viewing angles alongside a contour plot. 33 [PITH_FULL_IMAGE:figures/full_fig_p033_8.png] view at source ↗
read the original abstract

We propose a Bayesian framework for planning simple step-stress accelerated life tests when items are subject to two independent competing failure modes We assume that the competing risks are independent, with lifetimes following Weibull distributions, and adopt the cumulative exposure model with a log-linear stress-life relationship to connect failure time distributions across stress levels. The optimality criterion is the preposterior variance of the $p$-th quantile of the lifetime distribution at use stress, evaluated without reliance on asymptotic approximations, making the methodology valid regardless of sample size. Building on the idea of quantile-based reparametrisation used in single-mode ALT \citep{zhang2006bayesian}, we extend this approach to the competing risks setting by reparametrising the model parameters for each failure mode to physically interpretable and approximately independent quantities, making it possible to elicit priors directly from engineering knowledge of device behaviour. Posterior inference is carried out using the No-U-Turn Sampler implemented in Stan, and the optimal design is located via Monte Carlo simulation over a grid of candidate designs. The methodology is illustrated on a real step-stress dataset for a solar lighting device subject to capacitor and controller failure modes. A comprehensive sensitivity analysis with respect to the quantile probability, the lower stress level, the prior hyperparameter specification, and the sample size shows that the optimal stress-change time is moderately sensitive to these inputs while the optimal lower stress level consistently favours operation close to use conditions, a finding that holds across all prior specifications considered.

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 / 3 minor

Summary. The paper proposes a Bayesian framework for planning simple step-stress accelerated life tests (ALT) subject to two independent competing failure modes. Lifetimes are modeled as independent Weibulls under the cumulative exposure model with log-linear stress-life relationships. The optimality criterion is the preposterior variance of the p-th quantile of the use-condition lifetime distribution, computed exactly via prior-predictive Monte Carlo simulation and NUTS sampling in Stan rather than asymptotic approximations. Model parameters are reparametrized in terms of quantiles to facilitate prior elicitation from engineering knowledge. The approach is demonstrated on a solar lighting device dataset with capacitor and controller failure modes, accompanied by sensitivity analyses on p, lower stress level, prior hyperparameters, and sample size.

Significance. If the Monte Carlo implementation is accurate, the work supplies a practical exact Bayesian design tool for small-sample ALT with competing risks, extending quantile reparametrization from single-mode settings and removing reliance on large-sample approximations that are often unreliable in reliability experiments. The use of Stan for reproducible sampling and the explicit sensitivity analysis are strengths that enhance applicability in engineering contexts where sample sizes are limited.

major comments (2)
  1. [§4] §4 (Optimality criterion and computation): the preposterior variance is defined via Monte Carlo averaging over simulated data realizations, but the manuscript does not specify the exact number of outer Monte Carlo replications, the inner posterior sample size per replication, or convergence diagnostics for the NUTS sampler; these details are load-bearing for the claim that the criterion is evaluated 'exactly' and without asymptotic approximations.
  2. [§5] §5 (Numerical optimization): the optimal design is located by grid search over candidate (stress-change time, lower stress) pairs, yet no analysis is provided on grid resolution, potential for missing the global minimum, or comparison against a continuous optimizer; this directly affects the reliability of the reported optimal designs and sensitivity conclusions.
minor comments (3)
  1. [Abstract / §6] The abstract states that the optimal lower stress level 'consistently favours operation close to use conditions' across all priors, but the corresponding table or figure does not report the actual preposterior variance values at the boundary points, making it difficult to judge the magnitude of the improvement.
  2. [§3] Notation for the reparametrized quantile parameters (e.g., t_p,1 and t_p,2 for each failure mode) is introduced without an explicit mapping back to the original Weibull scale and shape parameters in one place; a single equation or table would improve clarity.
  3. [§6] The sensitivity analysis varies the quantile probability p but does not report how the elicited prior means and variances change with p, which would help readers reproduce the prior specifications.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments and for recognizing the practical utility of our exact Bayesian design framework. We address each major comment below and will incorporate the suggested clarifications into the revised manuscript.

read point-by-point responses

  1. Referee: [§4] §4 (Optimality criterion and computation): the preposterior variance is defined via Monte Carlo averaging over simulated data realizations, but the manuscript does not specify the exact number of outer Monte Carlo replications, the inner posterior sample size per replication, or convergence diagnostics for the NUTS sampler; these details are load-bearing for the claim that the criterion is evaluated 'exactly' and without asymptotic approximations.

    Authors: We agree that these implementation details are necessary to substantiate the exact, non-asymptotic nature of the preposterior variance computation. In the revised manuscript we will add a new paragraph in §4 that explicitly states the number of outer Monte Carlo replications performed, the number of post-warmup posterior draws retained from each NUTS run, and the convergence diagnostics (R-hat < 1.01 and minimum effective sample size) that were monitored across all replications. These additions will allow readers to verify the stability of the Monte Carlo estimate without altering the methodological claims. revision: yes

  2. Referee: [§5] §5 (Numerical optimization): the optimal design is located by grid search over candidate (stress-change time, lower stress) pairs, yet no analysis is provided on grid resolution, potential for missing the global minimum, or comparison against a continuous optimizer; this directly affects the reliability of the reported optimal designs and sensitivity conclusions.

    Authors: We accept that additional justification of the grid-search procedure is warranted. The revised §5 will report the exact grid spacing used for stress-change time and lower stress level, together with a supplementary check in which the grid is locally refined around the reported optimum to confirm that no superior design was overlooked. We will also note that the design variables are subject to engineering constraints that naturally favor a discrete search; a brief comparison with a derivative-free continuous optimizer will be included as a robustness check. These changes will strengthen confidence in the reported optima while preserving the computational simplicity of the original approach. revision: yes

Circularity Check

0 steps flagged

No significant circularity; preposterior variance criterion is simulation-defined and independent of fitted inputs

full rationale

The paper's core optimality criterion is the preposterior variance of the use-stress p-quantile, obtained by Monte Carlo simulation over prior-predictive data realizations and NUTS sampling in Stan. This quantity is defined directly from the prior and the likelihood without any reduction to a fitted parameter or self-referential equation within the paper. The quantile reparametrization extends a cited external method (Zhang et al. 2006) but does not smuggle an ansatz or rely on self-citation for uniqueness. Model assumptions (independent Weibulls, cumulative exposure, log-linear link) are stated explicitly as modeling choices rather than derived results. No step equates a prediction to its own input by construction, and the numerical grid search for the design is a direct evaluation of the stated criterion. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 3 axioms · 0 invented entities

The framework rests on standard domain assumptions from the ALT literature plus computational choices; no new entities are introduced.

free parameters (2)
  • prior hyperparameters
    Elicited from engineering knowledge of device behaviour and varied in sensitivity analysis.
  • quantile probability p
    Selected as part of the optimality criterion and examined in sensitivity checks.
axioms (3)
  • domain assumption Competing risks are independent
    Explicitly stated for the two failure modes.
  • domain assumption Lifetimes follow Weibull distributions
    Assumed for each failure mode.
  • domain assumption Cumulative exposure model with log-linear stress-life relationship
    Used to link distributions across stress levels.

pith-pipeline@v0.9.0 · 5556 in / 1531 out tokens · 79268 ms · 2026-05-10T17:01:12.968147+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. A Finite Mixture Failure-rate based Heterogeneous Step-stress Accelerated Life Testing (h-SSALT) Model

    stat.ME 2026-04 unverdicted novelty 6.0

    A finite mixture failure-rate based heterogeneous step-stress accelerated life testing model with Weibull Type-II censored data, EM estimation, and simulations showing systematic bias from ignoring heterogeneity.

  2. A Finite Mixture Failure-rate based Heterogeneous Step-stress Accelerated Life Testing (h-SSALT) Model

    stat.ME 2026-04 unverdicted novelty 6.0

    Proposes a failure-rate based heterogeneous SSALT model with finite mixtures of Weibull distributions for Type-II censored data, estimated via EM algorithm, as a generalization of prior cumulative exposure models.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages · cited by 1 Pith paper

  1. [1]

    Bhattacharyya, G. K. and Z. Soejoeti (1989). A tampered failure rate model for step-stress accelerated life test. Communications in Statistics-Theory and Methods\/ 18\/ (5), 1627--1643

    work page 1989

  2. [2]

    Bhattacharyya, R., B. N. Saha, G. G. Far \'i as, and N. Balakrishnan (2020). Multi-criteria-based optimal life-testing plans under hybrid censoring scheme. Test\/ 29\/ (2), 430--453

    work page 2020

  3. [3]

    Gelman, M

    Carpenter, B., A. Gelman, M. D. Hoffman, D. Lee, B. Goodrich, M. Betancourt, M. A. Brubaker, J. Guo, P. Li, and A. Riddell (2017). Stan: A probabilistic programming language. Journal of statistical software\/ 76 , 1--32

    work page 2017

  4. [4]

    Chaloner, K. and K. Larntz (1992). B ayesian design for accelerated life testing. Journal of Statistical Planning and Inference\/ 33\/ (2), 245--259

    work page 1992

  5. [5]

    De Groot , M. H. and P. K. Goel (1979). B ayesian estimation and optimal designs in partially accelerated life testing. Naval Research Logistics Quarterly\/ 26\/ (2), 223--235

    work page 1979

  6. [6]

    Ehrgott, M. (2005). Multicriteria Optimization\/ (2nd ed.), Volume 291. Berlin: Springer

    work page 2005

  7. [7]

    Erkanli, A. and R. Soyer (2000). Simulation-based designs for accelerated life tests. Journal of statistical planning and inference\/ 90\/ (2), 335--348

    work page 2000

  8. [8]

    and Y.-F

    Fan, T.-H. and Y.-F. Wang (2021). Comparison of optimal accelerated life tests with competing risks model under exponential distribution. Quality and Reliability Engineering International\/ 37\/ (3), 902--919

    work page 2021

  9. [9]

    Gelman, A., J. B. Carlin, H. S. Stern, D. B. Dunson, A. Vehtari, and D. B. Rubin (2013). B ayesian Data Analysis\/ (3 ed.). CRC Press

    work page 2013

  10. [10]

    Lee, and J

    Gelman, A., D. Lee, and J. Guo (2015). Stan: A probabilistic programming language for B ayesian inference and optimization. Journal of Educational and Behavioral Statistics\/ 40\/ (5), 530--543

    work page 2015

  11. [11]

    Ginebra, J. and A. Sen (1998). Minimax approach to accelerated life tests. IEEE transactions on reliability\/ 47\/ (3), 261--267

    work page 1998

  12. [12]

    and Y.-c

    Guan, Q. and Y.-c. Tang (2018). B ayesian planning of optimal step-stress accelerated life test for log-location-scale distributions. Acta Mathematicae Applicatae Sinica, English Series\/ 34\/ (1), 51--64

    work page 2018

  13. [13]

    Hamada, M. (2015). B ayesian analysis of step-stress accelerated life tests and its use in planning. Quality Engineering\/ 27\/ (3), 276--282

    work page 2015

  14. [14]

    Han, D. (2015). Time and cost constrained optimal designs of constant-stress and step-stress accelerated life tests. Reliability Engineering & System Safety\/ 140 , 1--14

    work page 2015

  15. [15]

    Han, D. and D. Kundu (2015). Inference for a step-stress model with competing risks for failure from the generalized exponential distribution under type- I censoring. IEEE Transactions on Reliability\/ 64\/ (1), 31--43

    work page 2015

  16. [16]

    Kateri, M. and U. Kamps (2015). Inference in step-stress models based on failure rates. Statistical Papers\/ 56 , 639--660

    work page 2015

  17. [17]

    Liu, X. and W. S. Qiu (2011). Modeling and planning of step-stress accelerated life tests with independent competing risks. IEEE Transactions on Reliability\/ 60\/ (4), 712--720

    work page 2011

  18. [18]

    and L.-C

    Liu, X. and L.-C. Tang (2010). Planning sequential constant-stress accelerated life tests with stepwise loaded auxiliary acceleration factor. Journal of Statistical Planning and Inference\/ 140\/ (7), 1968--1985

    work page 2010

  19. [19]

    Meeker, W. Q. and L. A. Escobar (1998). Statistical methods for reliability data . probability and statistics: Applied probability and statistics 314. New York, NY: John Wiley

    work page 1998

  20. [20]

    Nelson, W. B. (1980). Accelerated life testing: step-stress models and data analysis. IEEE Transactions on Reliability\/ 29\/ (2), 103--108

    work page 1980

  21. [21]

    Nelson, W. B. (1990). Accelerated Life Testing: Statistical Models, Test Plans and Data Analysis . New York: Wiley

    work page 1990

  22. [22]

    Pascual, F. (2007). Accelerated life test planning with independent W eibull competing risks with known shape parameter. IEEE Transactions on Reliability\/ 56\/ (1), 85--93

    work page 2007

  23. [23]

    Pascual, F. (2008). Accelerated life test planning with independent W eibull competing risks. IEEE transactions on reliability\/ 57\/ (3), 435--444

    work page 2008

  24. [24]

    Mitra, and D

    Prajapat, K., S. Mitra, and D. Kundu (2026). On constrained multicriteria-based planning for a two-parameter exponential distribution. Quality and Reliability Engineering International\/ 42\/ (1), 11--23

    work page 2026

  25. [25]

    Roy, S. (2018). B ayesian accelerated life test plans for series systems with W eibull component lifetimes. Applied Mathematical Modelling\/ 62 , 383--403

    work page 2018

  26. [26]

    Roy, S. and C. Mukhopadhyay (2016). B ayesian D -optimal accelerated life test plans for series systems with competing exponential causes of failure. Journal of Applied Statistics\/ 43\/ (8), 1477--1493

    work page 2016

  27. [27]

    Gupta, and D

    Samanta, D., A. Gupta, and D. Kundu (2019). Analysis of W eibull step-stress model in presence of competing risk. IEEE Transactions on Reliability\/ 68\/ (2), 420--438

    work page 2019

  28. [28]

    Singpurwalla, N. D. (1988). An interactive PC -based procedure for reliability assessment incorporating expert opinion and survival data. Journal of the American Statistical Association\/ 83\/ (401), 43--51

    work page 1988

  29. [29]

    Singpurwalla, N. D. (2006). Reliability and risk: a B ayesian perspective . England, UK: John Wiley & Sons

    work page 2006

  30. [30]

    Stephens, M. A. (1974). EDF statistics for goodness of fit and some comparisons. Journal of the American Statistical Association\/ 69\/ (347), 730--737

    work page 1974

  31. [31]

    Xu, A. and Y. Tang (2015a). A B ayesian method for planning accelerated life testing. IEEE Transactions on Reliability\/ 64\/ (4), 1383--1392

    work page

  32. [32]

    Xu, A. and Y. Tang (2015b). Reference optimality criterion for planning accelerated life testing. Journal of Statistical Planning and Inference\/ 167 , 14--26

    work page

  33. [33]

    Xu, A., Y. Tang, Q. Guan, and X. Lian (2016). Planning simple step-stress accelerated life tests using reference optimality criterion. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability\/ 230\/ (1), 85--92

    work page 2016

  34. [34]

    You, D. and H. Pham (2017). Self-adaptive stress accelerated life testing scheme. Journal of the Brazilian Society of Mechanical Sciences and Engineering\/ 39 , 2095--2103

    work page 2017

  35. [35]

    Yuan, T. and X. Liu (2011). B ayesian planning of optimal step-stress accelerated life test. In 2011 Proceedings-Annual Reliability and Maintainability Symposium , pp.\ 1--6. IEEE

    work page 2011

  36. [36]

    Liu, and W

    Yuan, T., X. Liu, and W. Kuo (2011). Planning simple step-stress accelerated life tests using B ayesian methods. IEEE Transactions on Reliability\/ 61\/ (1), 254--263

    work page 2011

  37. [37]

    Zhang, Y. and W. Q. Meeker (2005). B ayesian life test planning for the W eibull distribution with given shape parameter. Metrika\/ 61 , 237--249

    work page 2005

  38. [38]

    Zhang, Y. and W. Q. Meeker (2006). B ayesian methods for planning accelerated life tests. Technometrics\/ 48\/ (1), 49--60

    work page 2006

  39. [39]

    Zhou, Y., Z. Lu, Y. Shi, and K. Cheng (2020). B ayesian optimum accelerated life test plans based on quantile regression. Communications in Statistics-Simulation and Computation\/ 49\/ (9), 2402--2418

    work page 2020