pith. sign in

arxiv: 2008.06475 · v1 · submitted 2020-08-14 · 📊 stat.ME

A Maximin Φ_(p)-Efficient Design for Multivariate GLM

Yiou Li , Lulu Kang , Xinwei Deng This is my paper

Pith reviewed 2026-05-24 14:05 UTC · model grok-4.3

classification 📊 stat.ME
keywords maximin designΦ_p-efficiencymultivariate GLMoptimal experimental designmodel uncertaintyrobust designconvergent algorithm
0
0 comments X

The pith

A maximin Φ_p-efficient design maximizes the lowest efficiency across uncertain link functions, predictors, and parameters in multivariate GLMs.

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

Designs for generalized linear models lose efficiency when the link function, predictors, or parameter values are misspecified. Standard approaches either assume a fixed model or average over priors. This paper defines a maximin Φ_p-efficient criterion that selects the design maximizing the smallest Φ_p-efficiency over a set of possible model specifications. It then gives an algorithm that constructs such designs and proves the algorithm converges. Numerical examples show the resulting designs maintain good efficiency when the assumed model changes.

Core claim

The Mm-Φ_p design is obtained by maximizing, over all candidate designs, the minimum Φ_p-efficiency computed across a collection of possible link functions, predictor sets, and regression coefficient values; an algorithm exploiting monotonicity and continuity properties of this criterion constructs the design with guaranteed convergence.

What carries the argument

The Maximin Φ_p-Efficient (Mm-Φ_p) criterion, which replaces the usual Φ_p-optimality objective with the worst-case Φ_p-efficiency taken over model uncertainties.

If this is right

  • The constructed design guarantees a positive lower bound on Φ_p-efficiency no matter which model in the uncertainty set is true.
  • The algorithm terminates in finite steps under the stated theoretical conditions.
  • Performance can be checked by evaluating the minimum efficiency over the uncertainty set used in construction.
  • The same maximin formulation applies directly to any Φ_p criterion and to multivariate responses.

Where Pith is reading between the lines

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

  • The method supplies a deterministic robust alternative when prior distributions for Bayesian design are hard to elicit.
  • It could be applied to other optimality criteria such as D- or A-optimality by substituting the corresponding efficiency measure.
  • Extensions to continuous uncertainty sets would require replacing the inner minimization with a suitable numerical search.

Load-bearing premise

A well-defined, compact set of model uncertainties exists such that the minimum Φ_p-efficiency is attained and the algorithm's convergence properties apply.

What would settle it

An explicit set of link functions and parameter values for which the algorithm returns a design whose realized minimum Φ_p-efficiency is lower than that of a standard locally Φ_p-optimal design.

Figures

Figures reproduced from arXiv: 2008.06475 by Lulu Kang, Xinwei Deng, Yiou Li.

Figure 1
Figure 1. Figure 1: Boxplot of Worse-Case A- and D-Efficiency of Mm-Φ [PITH_FULL_IMAGE:figures/full_fig_p028_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Mm-Φp Design, Eff-Compromise Design, Centroid Optimal Design and Bayesian Optimal Design [PITH_FULL_IMAGE:figures/full_fig_p032_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Boxplot of A- and D-Efficiency of Four Designs at 10,000 Sam [PITH_FULL_IMAGE:figures/full_fig_p032_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: LEA(ξ (r) ,M′ ) of the r-th iteration in Algorithm 1. 4.3 Potato Packing Example We consider a real-world example, the potato packing example in Woods et al. (2006), to further evaluate the proposed Mm-Φp design. The experiment contains d = 3 quantitative variables - vitamin concentration in the prepackaging dip and the amount of two kinds of gas in the packing atmosphere. The response is binary representi… view at source ↗
Figure 5
Figure 5. Figure 5: Design Points of Mm-Φp Design and Compromise Designs [PITH_FULL_IMAGE:figures/full_fig_p052_5.png] view at source ↗
read the original abstract

Experimental designs for a generalized linear model (GLM) often depend on the specification of the model, including the link function, the predictors, and unknown parameters, such as the regression coefficients. To deal with uncertainties of these model specifications, it is important to construct optimal designs with high efficiency under such uncertainties. Existing methods such as Bayesian experimental designs often use prior distributions of model specifications to incorporate model uncertainties into the design criterion. Alternatively, one can obtain the design by optimizing the worst-case design efficiency with respect to uncertainties of model specifications. In this work, we propose a new Maximin $\Phi_p$-Efficient (or Mm-$\Phi_p$ for short) design which aims at maximizing the minimum $\Phi_p$-efficiency under model uncertainties. Based on the theoretical properties of the proposed criterion, we develop an efficient algorithm with sound convergence properties to construct the Mm-$\Phi_p$ design. The performance of the proposed Mm-$\Phi_p$ design is assessed through several numerical examples.

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

0 major / 2 minor

Summary. The manuscript proposes a Maximin Φ_p-Efficient (Mm-Φ_p) design criterion for multivariate generalized linear models that maximizes the minimum Φ_p-efficiency over uncertainties in link functions, predictors, and regression parameters. It develops an algorithm exploiting theoretical properties of the criterion (with claimed convergence guarantees) and assesses performance via numerical examples.

Significance. If the convergence properties and efficiency guarantees hold as stated, the work supplies a non-Bayesian robust-design alternative that directly optimizes worst-case Φ_p-efficiency; the explicit algorithm with convergence analysis is a concrete strength that could facilitate reproducible implementation.

minor comments (2)
  1. The abstract and introduction would benefit from a brief statement of the precise set of uncertainties (e.g., which link functions and predictor ranges) over which the maximin is taken, to make the scope of the numerical examples immediately clear.
  2. Notation for the multivariate GLM information matrix and the Φ_p criterion should be introduced with an explicit equation early in Section 2 to avoid ambiguity when the maximin is later defined.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the careful reading and positive assessment of the manuscript, including the recognition of the non-Bayesian robust-design approach and the explicit algorithm with convergence analysis. The recommendation for minor revision is noted. No specific major comments were raised in the report.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper defines the Mm-Φ_p criterion explicitly as the maximin of Φ_p-efficiency over a set of model uncertainties (link functions, predictors, parameters) and constructs an algorithm whose convergence relies on properties derived from that definition. No load-bearing step reduces a claimed prediction or result to a fitted input by construction, nor does any central claim rest on a self-citation chain that is itself unverified. The work is a direct proposal of a new design criterion and associated optimization procedure; the derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on the domain assumption that model uncertainties admit a well-defined minimax optimization over Φ_p efficiency; no free parameters or invented entities are identifiable from the abstract.

axioms (1)
  • domain assumption Model uncertainties (link functions, predictors, regression coefficients) can be represented such that a minimum Φ_p-efficiency is computable and the maximin design is well-defined for multivariate GLM.
    This premise is required for the proposed criterion to be applicable and is invoked in the abstract's description of the design goal.

pith-pipeline@v0.9.0 · 5701 in / 1315 out tokens · 21382 ms · 2026-05-24T14:05:43.456091+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

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

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.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.all := #1 'mid.sentence := #2 '...

    work page

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in " " * FUNCTION format....

    work page

  3. [3]

    Y., Parent, E., and Robert, C

    Amzal, B., Bois, F. Y., Parent, E., and Robert, C. P. (2006), Bayesian-Optimal Design via Interacting Particle Systems, Journal of the American Statistical Association, 101, 773--785

    work page 2006

  4. [4]

    Atkinson, A. C. (2014), Optimal Design, Wiley StatsRef: Statistics Reference Online, 1--17

    work page 2014

  5. [5]

    C., Donev, A

    Atkinson, A. C., Donev, A. N., and Tobias, R. D. (2006), Optimum Experimental Designs, with SAS, Oxford University Press

    work page 2006

  6. [6]

    Atkinson, A. C. and Woods, D. C. (2015), Designs for generalized linear models, Handbook of Design and Analysis of Experiments, 471--514

    work page 2015

  7. [7]

    and Dette, H

    Burghaus, I. and Dette, H. (2014), Optimal designs for nonlinear regression models with respect to non-informative priors, Journal of Statistical Planning and Inference, 154, 12--25

    work page 2014

  8. [8]

    Calafiore, G. C. and El Ghaoui, L. (2014), Optimization models, Cambridge university press

    work page 2014

  9. [9]

    (2015), Handbook of design and analysis of experiments, vol

    Dean, A., Morris, M., Stufken, J., and Bingham, D. (2015), Handbook of design and analysis of experiments, vol. 7, CRC Press

    work page 2015

  10. [10]

    Dror, H. A. and Steinberg, D. M. (2006), Robust experimental design for multivariate generalized linear models, Technometrics, 48, 520--529

    work page 2006

  11. [11]

    and Leonov, S

    Fedorov, V. and Leonov, S. (2013), Optimal Design for Nonlinear Response Models, Chapman & Hall/CRC Biostatistics Series

    work page 2013

  12. [12]

    (1974), On the Allocation of Linear Observations, Societas Scientiarum Fennica

    Fellman, J. (1974), On the Allocation of Linear Observations, Societas Scientiarum Fennica

    work page 1974

  13. [13]

    Fiacco, A. V. and Kortanek, K. O. (eds.) (1983), A Moment Inequality and Monotonicity of an Algorithm

    work page 1983

  14. [14]

    (2009), cvx users? guide, online: http://www

    Grant, M., Boyd, S., and Ye, Y. (2009), cvx users? guide, online: http://www. stanford. edu/\ boyd/software. html

    work page 2009

  15. [15]

    and Wong, W

    Imhof, L. and Wong, W. K. (2000), A graphical method for finding maximin efficiency designs, Biometrics, 56, 113--117

    work page 2000

  16. [16]

    I., Mukherjee, B., Sinha, B

    Khuri, A. I., Mukherjee, B., Sinha, B. K., and Ghosh, M. (2006), Design issues for generalized linear models: A review, Statistical Science, 376--399

    work page 2006

  17. [17]

    (1974), General equivalence theory for optimum designs (approximate theory), The annals of Statistics, 849--879

    Kiefer, J. (1974), General equivalence theory for optimum designs (approximate theory), The annals of Statistics, 849--879

    work page 1974

  18. [18]

    --- (1985), Optimal Design: Variation in Structure and Performance Under Change of Criterion, Biometrika, 62, 277--288

    work page 1985

  19. [19]

    and Wong, W

    King, J. and Wong, W. K. (1998), Optimal minimax designs for prediction in heteroscedastic models, Journal of Statistical Planning and Inference, 69, 371--383

    work page 1998

  20. [20]

    and Wong, W.-K

    King, J. and Wong, W.-K. (2000), Minimax D-optimal designs for the logistic model, Biometrics, 56, 1263--1267

    work page 2000

  21. [21]

    and Majumdar, D

    Li, G. and Majumdar, D. (2009), Some results on D-optimal designs for nonlinear models with applications, Biometrika, 96, 487--493

    work page 2009

  22. [22]

    and Deng, X

    Li, Y. and Deng, X. (2018), An Efficient Algorithm for Elastic I-optimal Design of Generalized Linear Models, arXiv preprint arXiv:1801.05861

    work page arXiv 2018

  23. [23]

    (1997), Maximin efficient designs Another view at D-optimality, Statistics & probability letters, 35, 109--114

    Schwabe, R. (1997), Maximin efficient designs Another view at D-optimality, Statistics & probability letters, 35, 109--114

    work page 1997

  24. [24]

    Silvey, S. D. (2013), Optimal design: an introduction to the theory for parameter estimation, vol. 1, Springer Science & Business Media

    work page 2013

  25. [25]

    D., Titterington, D

    Silvey, S. D., Titterington, D. M., and Torsney, B. (1978), An algorithm for optimal designs on a finite design space, Communications in Statistics - Theory and Methods, 14, 1379--1389

    work page 1978

  26. [26]

    Sitter, R. R. (1992), Robust designs for binary data, Biometrics, 1145--1155

    work page 1992

  27. [27]

    Sitter, R. R. and Torsney, B. (1995), D-Optimal Designs for Generalized Linear Models, in MODA4 --- Advances in Model-Oriented Data Analysis, eds. Kitsos, C. P. and M \"u ller, W. G., Heidelberg: Physica-Verlag HD, pp. 87--102

    work page 1995

  28. [28]

    Sobol', I. M. (1967), On the distribution of points in a cube and the approximate evaluation of integrals, Zhurnal Vychislitel'noi Matematiki i Matematicheskoi Fiziki, 7, 784--802

    work page 1967

  29. [29]

    Waite, T. W. and Woods, D. C. (2015), Designs for generalized linear models with random block effects via information matrix approximations , Biometrika, 102, 677--693

    work page 2015

  30. [30]

    (1992), A unified approach to the construction of minimax designs, Biometrika, 79, 611--619

    Wong, W.-K. (1992), A unified approach to the construction of minimax designs, Biometrika, 79, 611--619

    work page 1992

  31. [31]

    Wong, W. K. and Cook, R. D. (1993), Heteroscedastic G-optimal designs, Journal of the Royal Statistical Society: Series B (Methodological), 55, 871--880

    work page 1993

  32. [32]

    K., Yin, Y., and Zhou, J

    Wong, W. K., Yin, Y., and Zhou, J. (2019), Optimal Designs for Multi-Response Nonlinear Regression Models With Several Factors via Semidefinite Programming, Journal of Computational and Graphical Statistics, 28, 61--73

    work page 2019

  33. [33]

    Woods, D. C. and Lewis, S. M. (2011), Continuous optimal designs for generalized linear models under model uncertainty, Journal of Statistical Theory and Practice, 5, 137--145

    work page 2011

  34. [34]

    C., Lewis, S

    Woods, D. C., Lewis, S. M., Eccleston, J. A., and Russell, K. (2006), Designs for generalized linear models with several variables and model uncertainty, Technometrics, 48, 284--292

    work page 2006

  35. [35]

    C., Overstall, A

    Woods, D. C., Overstall, A. M., Adamou, M., and Waite, T. W. (2017), Bayesian design of experiments for generalized linear models and dimensional analysis with industrial and scientific application, Quality Engineering, 29, 91--103

    work page 2017

  36. [36]

    and Stufken, J

    Wu, H.-P. and Stufken, J. (2014), Locally _p -optimal designs for generalized linear models with a single-variable quadratic polynomial predictor , Biometrika, 101, 365--375

    work page 2014

  37. [37]

    (2013), On Optimal Designs for Nonlinear Models: A general and Efficient Algorithm, Journal of the American Statistical Association, 108, 1411--1420

    Yang, M., Biedermann, S., and Tang, E. (2013), On Optimal Designs for Nonlinear Models: A general and Efficient Algorithm, Journal of the American Statistical Association, 108, 1411--1420

    work page 2013

  38. [38]

    and Stufken, J

    Yang, M. and Stufken, J. (2009), Support points of locally optimal designs for nonlinear models with two parameters, The Annals of Statistics, 37, 518--541

    work page 2009

  39. [39]

    --- (2012), Identifying Locally Optimal Designs for Nonlinear Models: A Simple Extension with Profound Consequences, The Annals of Statistics, 40, 1665--1681

    work page 2012

  40. [40]

    (2011), Optimal designs for generalized linear models with multiple design variables, Statistica Sinica, 1415--1430

    Yang, M., Zhang, B., and Huang, S. (2011), Optimal designs for generalized linear models with multiple design variables, Statistica Sinica, 1415--1430

    work page 2011

  41. [41]

    (2010), Monotonic Convergence of a General Algorithm for Computing Optimal Designs, The Annals of Statistics, 38, 1593--1606

    Yu, Y. (2010), Monotonic Convergence of a General Algorithm for Computing Optimal Designs, The Annals of Statistics, 38, 1593--1606

    work page 2010