Pith. sign in

REVIEW 3 major objections 5 minor 39 references

R\'enyi-Induced Information Geometry and Hartigan's Prior Family

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Rényi’s geometry produces Hartigan’s priors exactly.

desk verdict Correct identification of Hartigan's priors with Renyi covolumes, wrapped in two overbroad claims that should be trimmed. read the letter →

arxiv 2506.12028 v1 pith:SARWGJM6 submitted 2025-05-22 math.ST stat.TH

Rebecca Maria Kuntz , Heinrich von Campe , Björn Malte Schäfer This is my paper · ORCID
classification math.STstat.TH MSC 62B1062F1553B12
keywords RényidivergenceinformationgeometrydualconnectionsFishermetricHartigan'spriorfamilyα-priorsJeffreysRényi-Laplace-Beltramioperator
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper derives the information geometry induced by the Rényi divergence: a metric tensor, dual connections, and dual Laplace–Beltrami operators. It shows that this Rényi geometry differs genuinely from Amari’s α-geometry, although both reduce to the Kullback–Leibler geometry in a limit. The central constructive result is that the covolumes parallel to the dual Rényi connections are canonical priors, and on exponential and mixture families they take the explicit forms $\rho^{n/2}(\det g_F)^{\rho}$ and $\rho^{n/2}(\det g_F)^{1-\rho}$. Comparing the parallelity equation with Hartigan’s defining equation for his parametrized prior family yields the paper’s main identification: Hartigan’s parameter $\alpha_H$ is exactly the Rényi order $\rho$. This matters because it turns a statistically motivated prior family into the canonical uniform priors of a divergence geometry and explains why the reparameterization $\alpha'=(1-\alpha)/2$ has special statistical status.

What carries the argument

The load-bearing object is the Rényi-parallel covolume: a smooth function $\operatorname{cov}(\theta)$ such that $\omega = \operatorname{cov}\, dx^1\wedge\cdots\wedge dx^n$ is parallel to one of the dual Rényi connections, which is equivalent to the first-order PDE $\partial_i\log\operatorname{cov} = \Gamma^{(\rho)}{}^{j}_{ji}$ for the primal connection and its dual analogue for $\rho^*$. On the exponential and mixture families the paper solves this PDE explicitly and obtains powers of $\det g_F$. The second central object is the reparameterization $\rho = (1-\alpha)/2$, which converts the Rényi covolume exponents $\rho$ and $1-\rho$ into the $\alpha$-prior exponents $(1-\alpha)/2$ and $(1+\alpha)/2$, thereby identifying the Rényi order parameter with Hartigan’s $\alpha_H$.

What would settle it

Choose a statistical family outside the exponential and mixture classes, for example a curved exponential family whose parameter manifold is compact, and test whether a smooth global solution of the Rényi parallelity PDE exists and whether Hartigan’s prior defined by equation (73) satisfies it; if any such family has no covolume or if the two definitions disagree, the claimed equality $\alpha_H = \rho$ fails beyond the families treated explicitly.

Watch

Extended reading notes

Core claim

Starting from the Rényi divergence $D_{\rho} = (\rho-1)^{-1}\log\int d\mu\, p^{\rho} (p')^{1-\rho}$, the paper applies the standard divergence construction to obtain $g^{(\rho)}_{ij} = \rho\, g^F_{ij}$, $\Gamma^{(\rho)}_{ijk} = \rho\, \Gamma^{(e)}_{ijk} + \rho^2 C_{ijk}$, and the dual $\Gamma^{(\rho*)}_{ijk} = \rho\, \Gamma^{(e)}_{ijk} + \rho(1-\rho) C_{ijk}$. Solving the parallelity conditions $\nabla^{(\rho)}\omega^{(\rho)}=0$ and $\nabla^{(\rho*)}\omega^{(\rho*)}=0$ on the exponential and mixture families yields the covolumes $\operatorname{cov}^{(\rho)}_e \propto (\det g_F)^{\rho}$ and $\operatorname{cov}^{(\rho*)}_e \propto (\det g_F)^{1-\rho}$, with the conformal factor $\rho^{n/2}$ absorbed into the volume form. Under the reparameterization $\rho = (1-\alpha)/2$, these coincide with the earlier $\alpha$-priors, and matching the general parallelity equation with Hartigan’s defining equation shows $\alpha_H = \rho$ for any suitable statistical family. The paper thus claims that Hartigan’s family is not merely analogous to the Rényi priors but is the same family.

Load-bearing premise

The whole identification rests on the assumption that every suitable statistical family admits a smooth, nowhere-vanishing covolume solving $\partial_i\log\operatorname{cov} = \Gamma^{(\rho)}{}^{j}_{ji}$; the paper constructs such solutions explicitly only for exponential and mixture families, and for an arbitrary family this partial differential equation need not have a global solution.

Editorial extensions

If this is right

  • Any statistical family with a global Rényi-parallel covolume inherits a canonical geometric prior, with Jeffreys prior recovered at $\rho=1/2$ and the Kullback–Leibler covolumes at $\rho=1$.
  • The special reparameterization $\alpha'=(1-\alpha)/2$, previously noticed through asymptotic estimator coincidences, is explained as the Rényi order parameter, giving Hartigan’s family a geometric rather than purely statistical justification.
  • The Rényi and $\alpha$ geometries are not the same geometry: the metric differs by the conformal factor $\rho$, the dual connections break the symmetry $\nabla^{(\rho*)} \neq \nabla^{(-\rho)}$, and no reparameterization makes their Laplace–Beltrami operators coincide.
  • On exponential and mixture families the dual covolumes swap under duality, $\omega^{(\rho*)}_e = \omega^{(\rho)}_m$ and $\omega^{(\rho)}_e = \omega^{(\rho*)}_m$, reflecting e/m duality at the level of priors.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the identification $\alpha_H = \rho$ holds generally, then Hartigan’s prior family inherits every geometric feature of the Rényi geometry, including its conformal scaling and its duality structure; one could look for statistical analogues of these geometric properties in Hartigan’s asymptotic results.
  • A concrete testable extension is to compute Rényi-parallel covolumes for curved exponential families or hierarchical models where the PDE cannot be solved in closed form, and to compare them with numerical solutions of Hartigan’s defining equation.
  • The conformal factor $\rho$ in the metric suggests that Rényi geometry may have interesting Weyl-geometric behaviour; if so, Hartigan’s priors would be tied to conformally invariant quantities, a connection the paper does not develop.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper derives the metric tensor, dual parametrized connections, dual Laplace-Beltrami operators, and parallel volume forms ('Renyi-priors') induced by the statistical Renyi divergence. It compares the resulting Renyi geometry with Amari's alpha-geometry, arguing that the two geometries are genuinely different in structure, and then identifies the Renyi-priors with Hartigan's parametrized prior family. The central identification is made by comparing Hartigan's defining PDE (Eq. 73) with the parallelity condition for a Renyi covolume (Eq. 75), yielding alpha_H = rho. The paper also uses this identification to give a geometric reason for Takeuchi and Amari's special reparameterization alpha' = (1-alpha)/2.

Significance. If taken with the necessary integrability qualification, the paper is a useful and largely correct contribution to information geometry. The central algebraic comparison is sound: the appendix's volume-form computations for the exponential and mixture families check out, including the KL limit (rho=1) and the Jeffreys/Bhattacharyya limit (rho=1/2), and no parameters are fitted. The identification of Hartigan's family with the Renyi-prior family is not circular, although it is essentially a direct comparison of two PDEs and therefore has a definitional flavor. The main weakness is that the claimed universality for 'any suitable family' is not established, since existence of a Renyi-parallel covolume is only demonstrated for e-flat and m-flat families. The paper's conceptual novelty is moderate but real: it gives a clean geometric explanation of the Takeuchi-Amari reparameterization.

major comments (3)
  1. [Section 4.4, Eq. (74)] The claim in the paragraph before Eq. (74) that the derivation 'is not limited to P_e or P_m, but holds true for statistical manifolds constructed from any suitable family' is unsupported. Equation (74) presupposes the existence of a smooth positive covolume solving partial_i log cov = Gamma(rho)^j_{ji}; for a general affine connection such a covolume exists locally only if the trace one-form tau_i = Gamma(rho)^j_{ji} is closed, and globally only under additional period conditions. The paper only verifies this for the exponential and mixture families in Appendix A, where the relevant connection is flat. For non-e-flat models such as curved exponential families, tau need not be a gradient, in which case no Renyi-parallel volume form exists. The identification with Hartigan's prior family should therefore be stated as conditional on existence/integrability of the covolume, or the universality claim should be restricted to families for which the PDE is solvable.
  2. [Section 3.3, Eq. (50)] The statement that 'there is no way to reparameterize the RLB-operator Delta(rho) to make it coincide with the alpha-LB-operator Delta(alpha)' is false as written: for rho=1 and alpha=-1 both operators equal Delta(m), and the paper itself notes that the KL limit is common. What is true is that no reparameterization maps the entire one-parameter families onto each other, because the coefficient of Delta(m) is fixed to 1 in Delta(rho) while it varies in Delta(alpha). Please revise Eq. (50) and the surrounding text to state this qualified claim.
  3. [Section 5, item 3] The concluding item states the reparameterization rho(alpha) = (1+alpha)/2, but the derivation in Section 4.4, Eq. (71) and item 5 uses rho = (1-alpha)/2. With rho = (1+alpha)/2 one obtains cov(rho)_e proportional to (det g_F)^((1+alpha)/2), which matches the alpha*-covolume, not the alpha-covolume. This sign inconsistency should be corrected, since it directly contradicts the paper's own main derivation.
minor comments (5)
  1. [Section 2.4, after Eq. (21)] The parameter domain is written as rho in R \ {-1,1}, but the divergence and the connections in Eq. (21)-(24) are parametrized by alpha; it should read alpha in R \ {-1,1}.
  2. [Equation (74)] The displayed formula for the parallelity condition is garbled: the wedge-product expression mixes indices j and k, and the final equality should read partial_i log cov = Gamma(rho)^k_{ki}. Please rewrite the derivation cleanly.
  3. [Table 1, Pm Dual Cov row] The entry writes cov^{B*}_m = cov^{LC}_e; since the LC covolume is the same for e and m up to the same factor, this is only an index typo, but it should be cov^{LC}_m for consistency.
  4. [Section 4.4, item 6] The phrase 'Amari and Takeuchi [4] recognize the reparameterization the alpha-priors' should read 'recognize the reparameterization of the alpha-priors'.
  5. [Section 5, item 3] The same sign error noted in the major comments appears again in the concluding item: the reparameterization should be rho(alpha) = (1-alpha)/2, not (1+alpha)/2, to match Eq. (71).

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the central Hartigan/Renyi identification is a derived equality between independently defined PDEs.

full rationale

The paper's derivation chain is self-contained on the circularity axis. The Renyi metric and dual connections (Eqs. 27-29) are computed from the Renyi divergence via Eguchi's standard formulas (Eqs. 10-12); the Renyi-priors are then defined geometrically as covolumes of volume forms parallel to the Renyi connections (Eqs. 60-61), not by Hartigan's equation. The main identification alpha_H = rho arises from comparing the derived PDE (Eq. 75) with Hartigan's defining equation (Eq. 73): after substituting the Renyi connection coefficients and the definitions of the e-connection and Amari-Chentsov tensor, the covolume equation has exactly the same form as Hartigan's family. This is a direct proof step, not a fitted input renamed as a prediction, and it does not presuppose the conclusion. The paper also explicitly derives the relation to Takeuchi and Amari's alpha-priors from the connection coefficients (item 5 of Sec. 4.4), so the reduction to a known reparameterization is acknowledged rather than hidden. No load-bearing self-citation occurs: the only reference involving an author of the present paper is [9], used for context on information geometry in cosmology. A separate rigor caveat, not a circularity, is that Eq. 74 assumes a global smooth covolume solving partial_i log cov = Gamma^(rho)^j_{ji}; Appendix A constructs solutions only for exponential and mixture families, so the claim in Sec. 4.4 that the derivation holds for 'any suitable family' is conditional on an unstated integrability condition. This limits the generality of the theorem but does not make the derivation circular. Score 1 reflects the absence of construction-forced equivalence; the final comparison is immediate once Eq. 75 is derived, but the derivation itself is independent.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central claim needs no fitted constants; the only continuous parameter rho labels the Renyi divergence and the priors. The mathematical inputs are Eguchi's construction, flatness of exponential and mixture families, existence of parallel covolumes, and Hartigan's defining PDE, all of which are standard or stated assumptions.

assumptions (5)
  • standard math Eguchi's divergence-induced geometry construction: metric g_D = -partial partial' D and dual connections from third derivatives of a divergence.
    Used in Section 2.2 to define all subsequent objects; requires the divergence to have vanishing first derivatives and positive definite mixed second derivative.
  • domain assumption The Renyi divergence D_rho for rho in R+ \ {1} satisfies Eguchi's conditions and is smooth in the parameters.
    Section 2.5 applies Eguchi's formulas directly to Eq 26; singular cases rho = 0, 1 are excluded.
  • domain assumption Exponential family is flat under nabla^(e) (Gamma^(e) = 0) and mixture family is flat under nabla^(m).
    Appendix A solves the parallel-volume PDE using this flatness; it is a standard textbook result but is an assumption about the model family.
  • domain assumption A smooth positive covolume solving partial_i log cov = Gamma^j_{ji} exists on the manifold considered.
    Eq 74 treats this PDE as well-posed for any suitable family; existence is only demonstrated explicitly for Pe and Pm in Appendix A.
  • domain assumption Hartigan's prior family is correctly defined by Eq 73.
    The comparison target is taken from Hartigan [28] and is assumed to characterize the alpha_H-parametrized family.

how reviews work

0 comments
Cite this review

Pith. "Pith review of R\'enyi-Induced Information Geometry and Hartigan's Prior Family." pith.science (2026). https://pith.science/paper/SARWGJM6

@misc{pith2026250612028,
  author       = {Pith},
  title        = {Pith review of: R\'enyi-Induced Information Geometry and Hartigan's Prior Family},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SARWGJM6}},
  note         = {Machine review of arXiv:2506.12028}
}
abstract

We derive the information geometry induced by the statistical R\'enyi divergence, namely its metric tensor, its dual parametrized connections, as well as its dual Laplacians. Based on these results, we demonstrate that the R\'enyi-geometry, though closely related, differs in structure from Amari's well-known $\alpha$-geometry. Subsequently, we derive the canonical uniform prior distributions for a statistical manifold endowed with a R\'enyi-geometry, namely the dual R\'enyi-covolumes. We find that the R\'enyi-priors can be made to coincide with Takeuchi and Amari's $\alpha$-priors by a reparameterization, which is itself of particular significance in statistics. Herewith, we demonstrate that Hartigan's parametrized ($\alpha_H$) family of priors is precisely the parametrized ($\rho$) family of R\'enyi-priors ($\alpha_H = \rho$).

Figures

Figures reproduced from arXiv: 2506.12028 by the authors.

Figure 1
Figure 1. Geometric structures are succesively built on a topological space. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 29 canonical work pages

  1. [1]

    Amari, Information Geometry and Its Applications, Vol

    S.-i. Amari, Information Geometry and Its Applications, Vol. 194 of Applied Mathematical Sciences, Springer Japan, 2016. doi:10.1007/978-4-431-55978-8 . URL http://link.springer.com/10.1007/978-4-431-55978-8

    doi:10.1007/978-4-431-55978-8 2016

  2. [2]

    H. Suyari, Generalization of shannon-khinchin axioms to nonextensive systems and the unique- ness theorem for the nonextensive entropy, IEEE Transactions on Information Theory 50 (8) (2004) 1783–1787. doi:10.1109/TIT.2004.831749

    work page arXiv 2004

  3. [3]

    V. M. Ili´ c, M. S. Stankovi´ c, Generalized shannon–khinchin axioms and uniqueness theorem for pseudo-additive entropies, Physica A: Statistical Mechanics and its Applications 411 (2014) 138–145. doi:10.1016/j.physa.2014.05.009. URL https://www.sciencedirect.com/science/article/pii/S0378437114003732

    work page doi:10.1016/j.physa.2014.05.009 2014

  4. [4]

    Takeuchi, S

    J. Takeuchi, S. Amari, α-parallel prior and its properties, IEEE Transactions on Information Theory 51 (3) (2005) 1011–1023. doi:10.1109/TIT.2004.842703

    work page arXiv 2005

  5. [5]

    D. C. de Souza, R. F. Vigelis, C. C. Cavalcante, Geometry induced by a generalization of r´ enyi divergence, Entropy 18 (11) (2016) 407. doi:10.3390/e18110407. URL https://www.mdpi.com/1099-4300/18/11/407

    work page doi:10.3390/e18110407 2016

  6. [6]

    Van Erven, P

    T. Van Erven, P. Harremos, R´ enyi divergence and kullback-leibler divergence, IEEE Transac- tions on Information Theory 60 (7) (2014) 3797–3820. 17

    work page 2014

  7. [7]

    Jiang, J

    R. Jiang, J. Tavakoli, Y. Zhao, Weyl prior and bayesian statistics, Entropy 22 (2020) 467. doi:10.3390/e22040467

    work page doi:10.3390/e22040467 2020

  8. [8]

    Nielsen, An elementary introduction to information geometry, Entropy 22 (10) (2020) 1100

    F. Nielsen, An elementary introduction to information geometry, Entropy 22 (10) (2020) 1100. doi:10.3390/e22101100. URL https://doi.org/10.3390%2Fe22101100

    doi:10.3390/e22101100 2020

Show all 39 references
  1. [9]

    Giesel, R

    E. Giesel, R. Reischke, B. M. Sch¨ afer, D. Chia, Information geometry in cosmological inference problems, JCAP 2021 (1) (2021) 005–005. arXiv:2005.01057[astro-ph,physics:gr-qc] , doi:10.1088/1475-7516/2021/01/005. URL http://arxiv.org/abs/2005.01057

    2021 arXiv

  2. [10]

    C. R. Rao, Information and the accuracy attainable in the estimation of statistical parameters, Bulletin of the Calcutta Mathematical Society 37 (1945) 81–91

    1945

  3. [11]

    R. A. Fisher, On the mathematical foundations of theoretical statistics, Philosophical transac- tions of the Royal Society of London. Series A, containing papers of a mathematical or physical character 222 (594-604) (1922) 309–368

    1922

  4. [12]

    Rattray, D

    M. Rattray, D. Saad, S.-i. Amari, Natural gradient descent for on-line learning, Physical review letters 81 (24) (1998) 5461

    1998

  5. [13]

    Girolami, B

    M. Girolami, B. Calderhead, Riemann manifold langevin and hamiltonian monte carlo meth- ods, Journal of the Royal Statistical Society Series B: Statistical Methodology 73 (2) (2011) 123–214. doi:10.1111/j.1467-9868.2010.00765.x. URL https://doi.org/10.1111/j.1467-9868.2010.00765.x

    2011

  6. [14]

    Efron, Defining the Curvature of a Statistical Problem (with Applications to Second Order Efficiency), The Annals of Statistics 3 (6) (1975) 1189 – 1242

    B. Efron, Defining the Curvature of a Statistical Problem (with Applications to Second Order Efficiency), The Annals of Statistics 3 (6) (1975) 1189 – 1242. doi:10.1214/aos/1176343282. URL https://doi.org/10.1214/aos/1176343282

    1975

  7. [15]

    Jeffreys, An invariant form for the prior probability in estimation problems, Proceedings of the Royal Society of London

    H. Jeffreys, An invariant form for the prior probability in estimation problems, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences 186 (1946) 453 – 461. URL https://api.semanticscholar.org/CorpusID:19490929

    1946

  8. [16]

    Nielsen, The many faces of information geometry, Notices of the American Mathematical Society 69 (2022) 36–45

    F. Nielsen, The many faces of information geometry, Notices of the American Mathematical Society 69 (2022) 36–45. doi:10.1090/noti2403

    2022 doi

  9. [17]

    Eguchi, Second order efficiency of minimum contrast estimators in a curved exponential family, The Annals of Statistics (1983) 793–803

    S. Eguchi, Second order efficiency of minimum contrast estimators in a curved exponential family, The Annals of Statistics (1983) 793–803

    1983

  10. [18]

    Bartelmann, General Relativity, Lecture Notes, Heidelberg University Publishing, 2019

    M. Bartelmann, General Relativity, Lecture Notes, Heidelberg University Publishing, 2019. doi:10.17885/heiup.534. URL https://heiup.uni-heidelberg.de/catalog/book/534

    2019 doi

  11. [19]

    B. M. Sch¨ afer, Tooltips for Theoretical Physics: Concepts of Modern Theoretical Physics, Scales and Mathematical Tools, Lecture Notes, Heidelberg University Publishing, 2022. doi: 10.17885/heiup.1059. URL https://heiup.uni-heidelberg.de/catalog/book/1059

    2022 doi

  12. [20]

    Kullback, R

    S. Kullback, R. A. Leibler, On information and sufficiency, The annals of mathematical statis- tics 22 (1) (1951) 79–86. 18

    1951

  13. [21]

    Chentsov, A

    N. Chentsov, A. M. Society, Statistical Decision Rules and Optimal Inference, Translations of mathematical monographs, American Mathematical Society, 1982. URL https://books.google.de/books?id=kTJJtQAACAAJ

    1982

  14. [22]

    Amari, Theory of information spaces: A differential geometrical foundation of statistics, Post RAAG Reports (1980)

    S.-i. Amari, Theory of information spaces: A differential geometrical foundation of statistics, Post RAAG Reports (1980)

    1980

  15. [23]

    R´ enyi, et al., On measures of information and entropy, in: Proceedings of the 4th Berkeley symposium on mathematics, statistics and probability, Vol

    A. R´ enyi, et al., On measures of information and entropy, in: Proceedings of the 4th Berkeley symposium on mathematics, statistics and probability, Vol. 1, 1961

    1961

  16. [24]

    Bhattacharyya, On a measure of divergence between two multinomial populations, Sankhy¯ a: The Indian Journal of Statistics (1933-1960) 7 (4) (1946) 401–406

    A. Bhattacharyya, On a measure of divergence between two multinomial populations, Sankhy¯ a: The Indian Journal of Statistics (1933-1960) 7 (4) (1946) 401–406. URL http://www.jstor.org/stable/25047882

    1946

  17. [25]

    Calin, C

    O. Calin, C. Udriste, Geometric Modeling in Probability and Statistics, Springer Cham, 2014. doi:10.1007/978-3-319-07779-6

    2014 doi

  18. [26]

    L. D. Brown, Admissible Estimators, Recurrent Diffusions, and Insoluble Boundary Value Problems, The Annals of Mathematical Statistics 42 (3) (1971) 855–903. doi:10.1214/aoms/ 1177693318. URL http://projecteuclid.org/euclid.aoms/1177693318

    1971

  19. [27]

    L. D. Brown, A Heuristic Method for Determining Admissibility of Estimators–With Applica- tions, The Annals of Statistics 7 (5) (1979) 960–994. URL http://www.jstor.org/stable/2958667

    1979

  20. [28]

    J. A. Hartigan, The maximum likelihood prior, The Annals of Statistics 26 (6) (1998) 2083 –

    1998

  21. [29]

    Komaki, A shrinkage predictive distribution for multivariate Normal observables, Biometrika 88 (3) (2001) 859–864

    F. Komaki, A shrinkage predictive distribution for multivariate Normal observables, Biometrika 88 (3) (2001) 859–864. doi:10.1093/biomet/88.3.859. URL https://doi.org/10.1093/biomet/88.3.859

    2001 doi

  22. [30]

    Amari, H

    S.-i. Amari, H. Nagaoka, Methods of information geometry, Vol. 191, American Mathematical Soc., 2000

    2000

  23. [31]

    Komaki, Shrinkage priors for bayesian prediction, The Annals of Statistics 34 (2) (Apr

    F. Komaki, Shrinkage priors for bayesian prediction, The Annals of Statistics 34 (2) (Apr. 2006). doi:10.1214/009053606000000010. URL http://dx.doi.org/10.1214/009053606000000010

    2006 doi

  24. [32]

    Komaki, Asymptotic Properties of Bayesian Predictive Densities When the Distributions of Data and Target Variables are Different, Bayesian Analysis 10 (1) (2015) 31 – 51

    F. Komaki, Asymptotic Properties of Bayesian Predictive Densities When the Distributions of Data and Target Variables are Different, Bayesian Analysis 10 (1) (2015) 31 – 51. doi: 10.1214/14-BA886. URL https://doi.org/10.1214/14-BA886

    2015 doi

  25. [33]

    R. E. Kass, L. A. Wasserman, The selection of prior distributions by formal rules, Journal of the American Statistical Association 91 (1996) 1343–1370. URL https://api.semanticscholar.org/CorpusID:53645083

    1996

  26. [34]

    Matsuzoe, Information geometry of bayesian statistics, AIP Conference Proceedings 1641 (1) (2015) 279–286

    H. Matsuzoe, Information geometry of bayesian statistics, AIP Conference Proceedings 1641 (1) (2015) 279–286. doi:10.1063/1.4905989. URL https://doi.org/10.1063/1.4905989 19

    2015 doi

  27. [35]

    Nakahara, Geometry, Topology and Physics, Second Edition, Graduate student series in physics, Taylor & Francis, 2003

    M. Nakahara, Geometry, Topology and Physics, Second Edition, Graduate student series in physics, Taylor & Francis, 2003. URL https://books.google.de/books?id=cH-XQB0Ex5wC

    2003

  28. [36]

    Lee, Introduction to Riemannian Manifolds, Graduate Texts in Mathematics, Springer In- ternational Publishing, 2019

    J. Lee, Introduction to Riemannian Manifolds, Graduate Texts in Mathematics, Springer In- ternational Publishing, 2019. URL https://books.google.de/books?id=UIPltQEACAAJ

    2019

  29. [37]

    Hartigan, Invariant prior distributions, The Annals of Mathematical Statistics 35 (2) (1964) 836–845

    J. Hartigan, Invariant prior distributions, The Annals of Mathematical Statistics 35 (2) (1964) 836–845. URL http://www.jstor.org/stable/2238537

    1964

  30. [38]

    J. A. Hartigan, The Asymptotically Unbiased Prior Distribution, The Annals of Mathematical Statistics 36 (4) (1965) 1137 – 1152. doi:10.1214/aoms/1177699988. URL https://doi.org/10.1214/aoms/1177699988 20

    1965

  31. [2103]

    URL https://doi.org/10.1214/aos/1024691462

    doi:10.1214/aos/1024691462. URL https://doi.org/10.1214/aos/1024691462

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.