pith. sign in

arxiv: 2605.19989 · v1 · pith:OCWJKDPHnew · submitted 2026-05-19 · 🧮 math.ST · cs.NA· math.NA· stat.TH

Error Bounds for Importance Sampling with Estimated Proposal Distributions

Cathrine Aeckerle-Willems , Ilja Klebanov , Simon Weissmann This is my paper

Pith reviewed 2026-05-20 04:12 UTC · model grok-4.3

classification 🧮 math.ST cs.NAmath.NAstat.TH
keywords importance samplingkernel density estimationnon-asymptotic error boundsMarkov chainsMonte Carlo methodsproposal distributiondefensive samplingself-normalized importance sampling
0
0 comments X

The pith

Error bounds for importance sampling with estimated proposals separate the Monte Carlo error from the proposal approximation error.

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

This paper derives non-asymptotic error bounds for importance sampling estimators that use a proposal distribution learned from data. The bounds distinguish the standard Monte Carlo error, which decreases like one over square root of the number of importance samples, from the error due to estimating the proposal density with a kernel density estimator based on an auxiliary sample. By providing rates for the kernel density estimator when samples come from a geometrically ergodic Markov chain, the results give concrete performance guarantees in terms of the sizes of the auxiliary and main samples. This matters for practitioners who routinely use data-driven proposals in importance sampling but previously had limited theory to rely on for understanding the combined errors.

Core claim

We address this gap by deriving non-asymptotic error bounds for standard, defensive, and self-normalized importance sampling estimators with random proposals. Our results separate the Monte Carlo error, scaling as n^{-1/2}, from the proposal approximation error measured through the mean integrated absolute and squared errors (MIAE and MISE) of the kernel density estimate. To obtain explicit convergence rates in (N,n), we establish MIAE and MISE bounds for KDEs constructed from geometrically ergodic Markov chains in stationary and non-stationary regimes. Combining these results yields quantitative guarantees for importance sampling with KDE-based proposals.

What carries the argument

Non-asymptotic error bounds that decompose total error into a Monte Carlo component of order n to the power minus one half and a proposal approximation component given by the mean integrated absolute or squared error of the kernel density estimate from the auxiliary Markov chain samples.

Load-bearing premise

The auxiliary samples for constructing the proposal estimate are produced by a geometrically ergodic Markov chain.

What would settle it

A simulation in which the observed error fails to separate into an n to the power of minus one half term and an integrated absolute or squared error term of the kernel density estimate would contradict the derived bounds.

Figures

Figures reproduced from arXiv: 2605.19989 by Cathrine Aeckerle-Willems, Ilja Klebanov, Simon Weissmann.

Figure 1.1
Figure 1.1. Figure 1.1: Two-stage importance sampling pipeline with estimated proposal distribution. Aux￾iliary samples are used to learn a proposal via kernel density estimation. Optionally, a defensive mixture stabilizes the proposal before importance sampling and estima￾tion of I(f). Importance sampling (IS; (Owen, 2013; Robert and Casella, 2004)) provides an alternative approach in which expectations under P are estimated u… view at source ↗
Figure 1
Figure 1. Figure 1 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 7
Figure 7. Figure 7 [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
Figure 7.1
Figure 7.1. Figure 7.1: Illustration of the target probability density function and proposal densities [PITH_FULL_IMAGE:figures/full_fig_p022_7_1.png] view at source ↗
Figure 7.2
Figure 7.2. Figure 7.2: Left: Performance of the defensive KDE IS estimator with fixed IS sample size n = 2000 depending on reference sample size N. Middle: Performance of the defensive KDE IS estimator with fixed reference sample size N = 200 depending on IS sample size n. Right: Performance of the defensive KDE IS estimator under joint scaling regimes. All plots show the MAE for different choices of δ (as functions of N) and … view at source ↗
read the original abstract

Importance sampling with data-driven proposal distributions is widely used in practice. A common workflow first generates an auxiliary sample of size $N$ from an approximation of the target distribution, constructs a density estimate $\hat q$ such as a kernel density estimator (KDE), and then draws $n$ importance samples from this learned proposal. Despite its practical relevance, the theoretical properties of this hierarchical procedure remain poorly understood, since classical importance sampling theory assumes a fixed proposal. We address this gap by deriving non-asymptotic error bounds for standard, defensive, and self-normalized importance sampling estimators with random proposals. Our results separate the Monte Carlo error, scaling as $n^{-1/2}$, from the proposal approximation error measured through the mean integrated absolute and squared errors (MIAE and MISE) of $\hat q$. To obtain explicit convergence rates in $(N,n)$, we establish MIAE and MISE bounds for KDEs constructed from geometrically ergodic Markov chains in stationary and non-stationary regimes. Combining these results yields quantitative guarantees for importance sampling with KDE-based proposals. Our theory provides practical guidance for selecting defensive mixture weights in a nonparametric importance sampling framework.

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 / 3 minor

Summary. The paper derives non-asymptotic error bounds for standard, defensive, and self-normalized importance sampling estimators that employ a random proposal distribution constructed via kernel density estimation from an auxiliary sample of size N. The auxiliary sample is drawn from a geometrically ergodic Markov chain, and the bounds separate the Monte Carlo error term of order n^{-1/2} from the proposal approximation error measured by the mean integrated absolute error (MIAE) and mean integrated squared error (MISE) of the KDE. Explicit convergence rates in (N,n) are obtained for both stationary and non-stationary regimes, yielding quantitative guarantees and guidance on defensive mixture weights.

Significance. If the central bounds hold, the manuscript fills a notable gap in the theory of importance sampling by providing rigorous non-asymptotic analysis for the common practical workflow of using data-driven proposals. The clean separation of Monte Carlo and approximation errors, together with explicit MIAE/MISE rates for KDEs under geometric ergodicity, supplies practical guidance on sample-size allocation and defensive weighting that is currently missing from the literature. The work correctly invokes standard mixing arguments and kernel-density results to obtain its rates.

minor comments (3)
  1. [§3.2] §3.2: the statement of the defensive estimator could be accompanied by a short remark clarifying how the mixture weight interacts with the random proposal to avoid potential reader confusion with the standard IS case.
  2. [Theorem 4.3] Theorem 4.3: the dependence of the leading constants on the geometric ergodicity rate is left implicit; a brief remark on how these constants scale with the mixing parameter would strengthen the practical utility of the rate statements.
  3. [Figure 1] Figure 1: the caption should explicitly state the values of N and n used in the simulation so that the plotted error curves can be directly compared to the derived bounds.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their supportive summary of the manuscript and for recommending minor revision. The assessment correctly identifies the separation of Monte Carlo and proposal approximation errors as a central contribution.

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained

full rationale

The paper derives non-asymptotic error bounds for importance sampling estimators with random KDE proposals by separating the Monte Carlo term (scaling as n^{-1/2}) from the proposal error measured in MIAE/MISE. These bounds are obtained by conditioning on an auxiliary sample from a geometrically ergodic Markov chain and invoking standard mixing and KDE convergence results for both stationary and non-stationary regimes. No step reduces by the paper's own equations to a fitted parameter, self-defined quantity, or load-bearing self-citation chain; the central claims rest on external, verifiable assumptions about ergodicity and kernel estimation that are not redefined within the work. The argument structure remains independent once the stated assumptions are granted.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard background results from Markov chain theory and nonparametric density estimation; no new free parameters or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption The auxiliary Markov chain is geometrically ergodic.
    Invoked to establish explicit MIAE and MISE bounds for the KDE in stationary and non-stationary regimes.

pith-pipeline@v0.9.0 · 5744 in / 1286 out tokens · 41224 ms · 2026-05-20T04:12:33.820096+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.

  • IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    We address this gap by deriving non-asymptotic error bounds for standard, defensive, and self-normalized importance sampling estimators with random proposals. Our results separate the Monte Carlo error, scaling as n^{-1/2}, from the proposal approximation error measured through the mean integrated absolute and squared errors (MIAE and MISE) of ˆq.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    To obtain explicit convergence rates in (N,n), we establish MIAE and MISE bounds for KDEs constructed from geometrically ergodic Markov chains in stationary and non-stationary regimes.

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

282 extracted references · 282 canonical work pages

  1. [1]

    The Annals of Statistics , number =

    Hanyuan Hang and Ingo Steinwart , title =. The Annals of Statistics , number =. 2017 , doi =

    work page 2017

  2. [2]

    The Annals of Statistics , number =

    Bin Yu , title =. The Annals of Statistics , number =. 1993 , doi =

    work page 1993

  3. [3]

    The Annals of Statistics , pages=

    Data-driven bandwidth choice for density estimation based on dependent data , author=. The Annals of Statistics , pages=. 1990 , publisher=

    work page 1990

  4. [4]

    IEEE Transactions on Information Theory , volume=

    Recursive probability density estimation for weakly dependent stationary processes , author=. IEEE Transactions on Information Theory , volume=. 1986 , publisher=

    work page 1986

  5. [5]

    Journal of Time Series Analysis , volume=

    Nonparametric estimators for time series , author=. Journal of Time Series Analysis , volume=. 1983 , publisher=

    work page 1983

  6. [6]

    Handbook of computational statistics: Concepts and methods , pages=

    Multivariate density estimation and visualization , author=. Handbook of computational statistics: Concepts and methods , pages=. 2011 , publisher=

    work page 2011

  7. [7]

    Electronic Journal of Statistics , number =

    Daniel Rudolf and Bj. Electronic Journal of Statistics , number =. 2020 , doi =

    work page 2020

  8. [8]

    On the convergence of the

    Schillings, Claudia and Sprungk, Bj\". On the convergence of the. Numer. Math. , FJOURNAL =. 2020 , NUMBER =

    work page 2020

  9. [9]

    , TITLE =

    Neddermeyer, Jan C. , TITLE =. J. Amer. Statist. Assoc. , FJOURNAL =. 2009 , NUMBER =

    work page 2009

  10. [10]

    Journal of the American Statistical Association , volume=

    Local adaptive importance sampling for multivariate densities with strong nonlinear relationships , author=. Journal of the American Statistical Association , volume=. 1996 , publisher=

    work page 1996

  11. [11]

    2026 , issn =

    Kernel density estimation with a Markov chain Monte Carlo sample , journal =. 2026 , issn =. doi:10.1016/j.csda.2025.108271 , author =

    work page doi:10.1016/j.csda.2025.108271 2026

  12. [12]

    The Journal of the Operational Research Society , author =

    Discrete. The Journal of the Operational Research Society , author =. 1994 , pages =. doi:10.2307/2584023 , language =

    work page doi:10.2307/2584023 1994

  13. [13]

    and Law, Kody and Stuart, Andrew M

    Iglesias, Marco A. and Law, Kody and Stuart, Andrew M. , title=. Inverse Problems , volume=

    work page

  14. [14]

    Goodman, Jonathan and Weare, Jonathan , doi =. Comm. App. Math. and Comp. Sci. , volumne =. 2010 , title =

    work page 2010

  15. [15]

    Robust random walk-like

    Rudolf, Daniel and Sprungk, Bj. Robust random walk-like. arXiv:2202.12127 , year =

    work page arXiv

  16. [16]

    J. Andr. Bayesian Analysis , number =. 2010 , doi =

    work page 2010

  17. [17]

    Proceedings of the 30th International Conference on Neural Information Processing Systems , pages =

    Liu, Qiang and Wang, Dilin , title =. Proceedings of the 30th International Conference on Neural Information Processing Systems , pages =. 2016 , isbn =

    work page 2016

  18. [18]

    Pinnau and C

    R. Pinnau and C. Totzeck and O. Tse and S. Martin. A consensus-based model for global optimization and its mean-field limit. Mathematical Models & Methods in Applied Sciences. 2017. doi:10.1142/S0218202517400061

    work page doi:10.1142/s0218202517400061 2017

  19. [19]

    The E nsemble K alman Filter: theoretical formulation and practical implementation

    Evensen, Geir. The E nsemble K alman Filter: theoretical formulation and practical implementation. Ocean Dynamics. 2003. doi:10.1007/s10236-003-0036-9

    work page doi:10.1007/s10236-003-0036-9 2003

  20. [20]

    and Kroese, Dirk P

    Rubinstein, Reuven Y. and Kroese, Dirk P. , year =. Simulation and the

    work page

  21. [21]

    and Rubinstein, Reuven Y

    Kroese, Dirk P. and Rubinstein, Reuven Y. and Glynn, Peter W. , year =. The. Handbook of. doi:10.1016/B978-0-444-53859-8.00002-3 , pages =

    work page doi:10.1016/b978-0-444-53859-8.00002-3

  22. [22]

    INFORMS Journal on Computing , author =

    A. INFORMS Journal on Computing , author =. 2007 , pages =. doi:10.1287/ijoc.1060.0176 , language =

    work page doi:10.1287/ijoc.1060.0176 2007

  23. [23]

    Chan, Joshua C C and Glynn, Peter W and Kroese, Dirk P , pages =. A

    work page

  24. [24]

    Operations Research Letters , author =

    Convergence properties of the cross-entropy method for discrete optimization , volume =. Operations Research Letters , author =. 2007 , pages =. doi:10.1016/j.orl.2006.11.005 , abstract =

    work page doi:10.1016/j.orl.2006.11.005 2007

  25. [25]

    Dambreville, Frederic , pages =. Cross-

    work page

  26. [26]

    Rare event estimation for static models via cross-entropy and importance sampling , author =

    work page

  27. [27]

    SIAM Journal on Scientific Computing , author =

    Active subspace methods in theory and practice: applications to kriging surfaces , volume =. SIAM Journal on Scientific Computing , author =. 2014 , note =. doi:10.1137/130916138 , abstract =

    work page doi:10.1137/130916138 2014

  28. [28]

    The Annals of Applied Probability , author =

    The sample size required in importance sampling , volume =. The Annals of Applied Probability , author =. 2018 , keywords =. doi:10.1214/17-AAP1326 , language =

    work page doi:10.1214/17-aap1326 2018

  29. [29]

    , year =

    Chatfield, Christopher and Collins, Alexander J. , year =. Introduction to. doi:10.1007/978-1-4899-3184-9 , keywords =

    work page doi:10.1007/978-1-4899-3184-9

  30. [30]

    Tempered,

    Aufort, Grégoire and Pudlo, Pierre and Burgarella, Denis , month = jun, year =. Tempered,

    work page

  31. [31]

    Safety and Reliability , author =

    Improved cross entropy-based importance sampling for network reliability assessment , abstract =. Safety and Reliability , author =. 2021 , keywords =

    work page 2021

  32. [32]

    Reliability-oriented sensitivity analysis under probabilistic model uncertainty –

    Chabridon, Vincent , keywords =. Reliability-oriented sensitivity analysis under probabilistic model uncertainty –

    work page

  33. [33]

    Statistics and Computing , author =

    Sequential. Statistics and Computing , author =. 2012 , keywords =. doi:10.1007/s11222-011-9231-6 , abstract =

    work page doi:10.1007/s11222-011-9231-6 2012

  34. [34]

    Probabilistic Engineering Mechanics , author =

    Estimation of small failure probabilities in high dimensions by subset simulation , volume =. Probabilistic Engineering Mechanics , author =. 2001 , keywords =. doi:10.1016/S0266-8920(01)00019-4 , abstract =

    work page doi:10.1016/s0266-8920(01)00019-4 2001

  35. [35]

    Digital Signal Processing , author =

    Adaptive importance sampling in signal processing , volume =. Digital Signal Processing , author =. 2015 , keywords =. doi:10.1016/j.dsp.2015.05.014 , language =

    work page doi:10.1016/j.dsp.2015.05.014 2015

  36. [36]

    IEEE Signal Processing Magazine , author =

    Adaptive. IEEE Signal Processing Magazine , author =. 2017 , keywords =. doi:10.1109/MSP.2017.2699226 , language =

    work page doi:10.1109/msp.2017.2699226 2017

  37. [37]

    Journal of Computational and Graphical Statistics , author =

    Population monte carlo , volume =. Journal of Computational and Graphical Statistics , author =. 2004 , note =. doi:10.1198/106186004X12803 , number =

    work page doi:10.1198/106186004x12803 2004

  38. [38]

    Reliability Engineering & System Safety , author =

    A new uncertainty importance measure , volume =. Reliability Engineering & System Safety , author =. 2007 , keywords =. doi:10.1016/j.ress.2006.04.015 , abstract =

    work page doi:10.1016/j.ress.2006.04.015 2007

  39. [39]

    , year =

    Bishop, Christopher M. , year =. Pattern recognition and machine learning , isbn =

    work page

  40. [40]

    Curse-of-dimensionality revisited:

    Bengtsson, Thomas and Bickel, Peter and Li, Bo , year =. Curse-of-dimensionality revisited:. Institute of. doi:10.1214/193940307000000518 , keywords =

    work page doi:10.1214/193940307000000518

  41. [41]

    and Papaspiliopoulos, O

    Agapiou, S. and Papaspiliopoulos, O. and Sanz-Alonso, D. and Stuart, A. M. , month = jan, year =. Importance

    work page

  42. [42]

    Structural Safety , author =

    Important sampling in high dimensions , volume =. Structural Safety , author =. 2003 , keywords =. doi:10.1016/S0167-4730(02)00047-4 , abstract =

    work page doi:10.1016/s0167-4730(02)00047-4 2003

  43. [43]

    Au, Siu-Kui , year =. On the

    work page

  44. [44]

    Safe and

    Owen, Art and Zhou, Yi , year =. Safe and

    work page

  45. [45]

    Mathematics of Computation , author =

    On some inequalities for the gamma and psi functions , volume =. Mathematics of Computation , author =. 1997 , keywords =. doi:10.1090/S0025-5718-97-00807-7 , abstract =

    work page doi:10.1090/s0025-5718-97-00807-7 1997

  46. [46]

    Stochastics and Stochastic Reports , author =

    Importance. Stochastics and Stochastic Reports , author =. 2004 , pages =. doi:10.1080/10451120410001733845 , language =

    work page doi:10.1080/10451120410001733845 2004

  47. [47]

    The Annals of Statistics , author =

    Empirical-likelihood-based confidence interval for the mean with a heavy-tailed distribution , volume =. The Annals of Statistics , author =. doi:10.1214/009053604000000328 , language =

    work page doi:10.1214/009053604000000328

  48. [48]

    Physical Review D , author =

    Estimation of cosmological parameters using adaptive importance sampling , volume =. Physical Review D , author =. 2018 , note =. doi:10.1103/PhysRevD.80.023507 , abstract =

    work page doi:10.1103/physrevd.80.023507 2018

  49. [49]

    Curse-of-dimensionality revisited:

    Li, Bo and Bengtsson, Thomas and Bickel, Peter , pages =. Curse-of-dimensionality revisited:

    work page

  50. [50]

    Nonasymptotic bounds for suboptimal importance sampling , url =

    Hartmann, Carsten and Richter, Lorenz , month = feb, year =. Nonasymptotic bounds for suboptimal importance sampling , url =

    work page

  51. [51]

    and Mannor, Shie and Rubinstein, Reuven Y

    A. Annals of Operations Research , author =. 2005 , pages =. doi:10.1007/s10479-005-5724-z , abstract =

    work page doi:10.1007/s10479-005-5724-z 2005

  52. [52]

    Importance

    Sanz-Alonso, Daniel , month = aug, year =. Importance

    work page

  53. [53]

    Journal of the American statistical association , author =

    Sequential imputations and. Journal of the American statistical association , author =. 1994 , keywords =

    work page 1994

  54. [54]

    Sequential importance sampling for structural reliability analysis , journal =

    Iason Papaioannou and Costas Papadimitriou and Daniel Straub , keywords =. Sequential importance sampling for structural reliability analysis , journal =. 2016 , issn =. doi:https://doi.org/10.1016/j.strusafe.2016.06.002 , url =

    work page doi:10.1016/j.strusafe.2016.06.002 2016

  55. [55]

    Insight from the

    Shadmi, Yonatan and Simatos, Florian , year =. Insight from the

    work page

  56. [56]

    and Kroese, Dirk P

    Rubinstein, Reuven Y. and Kroese, Dirk P. , TITLE =. 2004 , PAGES =

    work page 2004

  57. [57]

    2022 , eprint=

    Optimal projection to improve parametric importance sampling in high dimension , author=. 2022 , eprint=

    work page 2022

  58. [58]

    Annealed importance sampling , abstract =

    Neal, Radford M , year =. Annealed importance sampling , abstract =

    work page

  59. [59]

    Owen, Art , year =. Monte

    work page

  60. [60]

    National Bureau of Standards applied mathematics series , author =

    Estimation of particle transmission by random sampling , volume =. National Bureau of Standards applied mathematics series , author =. 1951 , note =

    work page 1951

  61. [61]

    A benchmark study on importance sampling techniques in structural reliability , abstract =

    Engelund, S and Rackwitz, R , year =. A benchmark study on importance sampling techniques in structural reliability , abstract =

    work page

  62. [62]

    Rubinstein, Reuven , year =. The

    work page

  63. [63]

    Certified dimension reduction in nonlinear

    Zahm, Olivier and Cui, Tiangang and Law, Kody and Spantini, Alessio and Marzouk, Youssef , month = jan, year =. Certified dimension reduction in nonlinear

    work page

  64. [64]

    Weighted

    Hesterberg, Timothy Classen , year =. Weighted

    work page

  65. [65]

    Expert Systems With Applications , author =

    Fight sample degeneracy and impoverishment in particle filters:. Expert Systems With Applications , author =. 2014 , pages =

    work page 2014

  66. [66]

    Recursive

    El-Laham, Yousef and Elvira, Victor and Bugallo, Monica , year =. Recursive

    work page

  67. [67]

    Hesterberg, Timothy Classen , year =

    work page

  68. [68]

    and Kent, John T

    Mardia, Kantilal V. and Kent, John T. and Bibby, John M. , year =. Multivariate analysis , isbn =

    work page

  69. [69]

    Moments for the

    Rosen, Dietrich Von , year =. Moments for the

    work page

  70. [70]

    Statistical Science , author =

    Generalized. Statistical Science , author =. doi:10.1214/18-STS668 , abstract =

    work page doi:10.1214/18-sts668

  71. [71]

    Journal of Computational and Graphical Statistics , author =

    Truncated. Journal of Computational and Graphical Statistics , author =. 2008 , pages =. doi:10.1198/106186008X320456 , language =

    work page doi:10.1198/106186008x320456 2008

  72. [72]

    and Casella, George , year =

    Robert, Christian P. and Casella, George , year =. Monte

    work page

  73. [73]

    , year =

    Liu, Jun S. , year =. Monte

    work page

  74. [74]

    M and Handscomb, D

    Hammersley, J. M and Handscomb, D. C , year =. Monte

    work page

  75. [75]

    Goertzel, Gerald , year =. A

    work page

  76. [76]

    Journal of Statistical Computation and Simulation , author =

    Adaptive importance sampling in monte carlo integration , volume =. Journal of Statistical Computation and Simulation , author =. 1992 , pages =. doi:10.1080/00949659208810398 , language =

    work page doi:10.1080/00949659208810398 1992

  77. [77]

    Econometrica , author =

    Bayesian. Econometrica , author =. 1989 , pages =. doi:10.2307/1913710 , language =

    work page doi:10.2307/1913710 1989

  78. [78]

    Econometrica , author =

    Bayesian. Econometrica , author =. 1978 , pages =. doi:10.2307/1913641 , language =

    work page doi:10.2307/1913641 1978

  79. [79]

    , year =

    On the. Annals of Operations Research , author =. 2005 , pages =. doi:10.1007/s10479-005-5731-0 , abstract =

    work page doi:10.1007/s10479-005-5731-0 2005

  80. [80]

    Journal of Statistical Planning and Inference , author =

    Random matrix theory in statistics:. Journal of Statistical Planning and Inference , author =. 2014 , pages =. doi:10.1016/j.jspi.2013.09.005 , abstract =

    work page doi:10.1016/j.jspi.2013.09.005 2014

Showing first 80 references.