Pith.

Instrumentation and Methods for Astrophysics · arXiv:2611.17231 v1 · submitted 12 Nov 2026 · 6 references

Estimating distances from parallaxes. VII. Computing the three-dimensional kinegeometric posterior with one-dimensional sampling

Coryn A.L. Bailer-Jones

The paper says

“When parallaxes are of poor quality, proper motions can improve stellar distance estimates using known constraints on stars' physical velocities.”

more of the abstract

Formally this may be achieved via the kinegeometric posterior, a three-dimensional (3D) posterior in distance and velocity for given measured parallax and 2D proper motion. This posterior is the product of a covariant Gaussian likelihood in the parallax and proper motion with a distance and velocity prior that can be constructed from a Galaxy model. In earlier work, I showed how to infer this posterior for stars in Gaia DR3 by explicitly sampling it with a Markov Chain Monte Carlo method. Here I show that this posterior is equivalent to a 1D posterior in distance combined with a sum of 2D Gaussian distributions in velocity. This 1D+2D representation is both faster and more accurate to sample, thereby making the computation of the kinegeometric posterior for a large set of sources much easier.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

Holds up. The 1D+2D reduction of the kinegeometric posterior is exact and correctly derived; a genuine methodological advance, with only the untested non-Gaussian velocity prior and a covariance typo in Eq. (7) as real soft spots. This line is wrong

4 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 6 references say.

4 minor objections

The paper says, at Section 3, Eq. (7)

Pith says: minor objection 1 of 4

The second argument of N_v should be k^2 r^2 Sigma_2, not k r Sigma_2. As printed, the covariance has incorrect dimensions (it is a covariance in velocity and must have units of velocity squared), and Eqs. (6), (8) and (10) consistently use k^2 r^2 Sigma_2. This is a display typo that does not affect the rest of the derivation, but the equation should be corrected.

The paper says, at Abstract and Section 1

Pith says: minor objection 2 of 4

The equivalence result is derived under a Gaussian velocity prior conditional on distance and direction. Section 3 states this condition, but the abstract's unqualified wording, 'this posterior is equivalent to...', could be misread as holding for any velocity prior. Add the qualifier to the abstract.

show 2 more

The paper says, at Section 3, Step 6

Pith says: minor objection 3 of 4

The sentence 'for a given proper motion, an increase in the distance can be compensated for by a decrease in the velocity' is incorrect: Eq. (1) gives v = k r mu, so at fixed proper motion the transverse velocity increases, not decreases, with distance. Rephrase to describe the positive r-v correlation implied by the likelihood.

The paper says, at Section 5, Fig. 3

Pith says: minor objection 4 of 4

Several ratio distributions have very large standard deviations (for example, vraMedKinegeo sigma about 5.2 and rvdecCorrKinegeo sigma about 1.9), so the quoted medians and standard deviations are dominated by outliers and disagree with the visual impression of the histograms. Report a robust scatter estimate, such as interpercentile ranges or MAD, and briefly comment on the sources of the outliers.

What would settle it

Recompute Eq. (7) directly from Eq. (6): the Gaussian in v has mean kr[mu - X_mu] and covariance k^2 r^2 Sigma_2, not kr Sigma_2; verify the implementation uses k^2 r^2 Sigma_2 in Eqs. (8)-(10). Then select one source with strong parallax-proper-motion correlation and compare 1D+2D posterior summaries against a dense-grid numerical marginalization of the 3D posterior; agreement within Monte Carlo error would confirm the corrected equation and the implementation.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up confidence high
  2. Skeptic partly agrees 3 min 19 s · 10,425 tokens read, 24,181 written
  3. Editor wrote the verdict above 18 s · 11,189 tokens read, 2,819 written
  4. Referee minor revision: 4 minor 2 min 21 s · 10,801 tokens read, 18,320 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

This paper is right. The core claim — that the 3D kinegeometric posterior factorizes exactly into a 1D marginal over distance times a conditional 2D Gaussian in velocity — holds under the stated Gaussian velocity prior. I checked the algebra in Section 3 and the change-of-variables step is correct: because the proper-motion likelihood's mean is linear in v at fixed r, the v-dependence collapses to a single Gaussian, and the marginalization over v in Step 4 leaves a clean 1D problem. The Monte Carlo representation of the velocity posterior as a sum of Gaussians in Step 5 is standard and sound.

What's genuinely new: paper VI sampled the 3D posterior directly with MCMC; this shows the 1D+2D equivalence, which was not previously demonstrated. That matters in practice. The numerical section is convincing where it needs to be: on 500 real Gaia DR3 stars, the medians of all distance and velocity quantiles and the three correlation coefficients agree with the old 3D MCMC to well within 1%, and the 15x speedup holds even when the 1D sampling is done with emcee rather than PINV. That's honest — the author admits emcee in 1D costs only 50% more, so the speedup is mainly from dimensionality reduction, as claimed.

Soft spots, in proportion. The Gaussian velocity prior is the load-bearing condition, and the paper states it openly in Section 3, Step 2; it is not a hidden assumption. But robustness to non-Gaussian velocity priors is untested, and that is the main boundary on applicability — if the true velocity field is multimodal or heavy-tailed, the conditional posteriors won't be Gaussian and the reduction won't hold exactly. The demonstration is a consistency check against the same prior, not an independent validation against truth, which is the right design for a claim about exactness, so I don't hold that against it. One concrete defect: Eq. (7) prints the covariance of the transformed velocity Gaussian as kr·Sigma2, but the consistent expression from Eqs. (6), (8), and (10) is k2 r2 Sigma2. It is a display typo, does not affect the derivation or conclusions, but should be corrected. The velocity-prior parameters come from paper VI's mock-catalogue calibration; that is cited external input, not circularity.

Who it's for: anyone doing distance or transverse-velocity inference from Gaia parallax and proper motion, and astrostatistics practitioners generally. It deserves a serious referee and, after the typo fix, publication. I'd take it to reading group.

The strongest objection

After checking the algebra in Section 3, I find no load-bearing objection to the central claim. Equations (6)-(10) correctly transform the conditional proper-motion likelihood into a Gaussian in v, complete the square with the 2D Gaussian velocity prior, and arrive at the conditional posterior N_v(mv, Sigma_v) with Q(r, varpi, mu) absorbing all v-independent terms. Equations (12b) and (13c) then give the marginal distance posterior and the Monte Carlo Gaussian-mixture representation of the marginal velocity posterior, so drawing r from g(r) and one v from each conditional Gaussian does generate exact samples from the 3D posterior. The Gaussian velocity prior is an explicit scope condition stated in Section 3, Step 2 ('By further adopting a Gaussian prior...'), not a hidden assumption; the derivation is exact under that model, and the absence of a non-Gaussian robustness test does not undermine the methodological claim. The only concrete defect I see is a typographical inconsistency in Eq. (7): the transformed Gaussian in v is printed with covariance kr*Sigma_2, whereas Eq. (6), Eq. (8), and Eq. (10) consistently use k^2 r^2 Sigma_2. The derivation's conclusion does not depend on the erroneous printed covariance, but the displayed equation should be corrected.

The reader's weakest-assumption identification points at the Gaussian velocity prior, and that is indeed the condition on which the closed-form 1D+2D representation rests. However, the paper states this condition explicitly rather than concealing it, and the algebraic reduction is exact under it. A non-Gaussian velocity prior would also misspecify the 3D MCMC reference method used in the comparison, so the lack of a robustness study is a practical scope limitation, not a flaw in the derivation or in the claimed equivalence for the stated model. The numerical comparison with the 3D MCMC on 500 Gaia DR3 sources is a reasonable sanity check, though the analytic derivation is the primary evidence. The one concrete internal issue is Eq. (7), which prints the transformed covariance as kr Sigma_2 instead of k^2 r^2 Sigma_2; the surrounding equations correct this, so it does not change the central conclusion. The paper's claims about speed and accuracy are empirical and plausible, and the acknowledged PINV interpolation and finite-sample limitations are not load-bearing for the methodological contribution. Verdict unchanged.

The referee's report

This is a useful and clean methodological contribution to the established distance-inference series. The 1D+2D factorization is nontrivial and exact under the stated assumptions; it replaces a 3D MCMC computation by a 1D numerical inversion plus Gaussian draws, which is simpler and less prone to convergence problems. The derivation is self-contained, and I checked Eqs. (6)-(13): the change of variables, the completing-the-square step, and the Monte Carlo mixture representation of the velocity marginal are all valid. The numerical demonstration is appropriately treated as a consistency check between two samplers using the same prior rather than as an astrophysical validation of the prior. The main scope condition, the Gaussian form of the velocity prior, is stated explicitly in the derivation, so it does not undermine the central claim.

Pith's reading of the paper (a machine's summary, not the authors')

This paper shows that the three-dimensional (distance and two velocity components) kinegeometric posterior can be reduced without approximation to a one-dimensional distance posterior plus a sum of two-dimensional Gaussian velocity posteriors. This analytic reduction replaces a slow, unreliable MCMC sampling of a 3D space with fast, accurate 1D sampling followed by direct draws from 2D Gaussians. If correct, the method makes kinegeometric distance and velocity estimates practical for the billions of stars in Gaia data releases, transforming a previously expensive calculation into a routine one.

What carries the argument. The linear mapping $v = k r \mu$ (with $k=4.740471\,\mathrm{km\,s^{-1}\,kpc^{-1}\,mas^{-1}\,yr}$) converts the 2D proper-motion likelihood conditional on parallax into a Gaussian in velocity. This Gaussian multiplies with the Gaussian velocity prior to yield another Gaussian in velocity via completing the square. The resulting 1D+2D factorization reduces the 3D posterior sampling to a 1D sampling of the distance posterior $g(r)$ (equation 12b) followed by draws from the 2D Gaussian conditional velocity posterior $N_v(m_v,\Sigma_v)$ (equation 10). The formalism is exact under the assumption of a Gaussian velocity prior.

Core claim. The joint posterior over distance and transverse velocity, given a parallax and two proper motions, factorizes into a product of a 1D distance posterior and a 2D conditional velocity posterior that is exactly a Gaussian. The key step is rewriting the 2D proper-motion likelihood conditional on parallax as a Gaussian in velocity (using the linear relation $v = k r \mu$), which then multiplies with the Gaussian velocity prior to produce another Gaussian in velocity. The marginal distance posterior becomes the geometric posterior multiplied by a scalar function $Q(r,\varpi,\mu)$ that depends only on distance and the data. The full 3D posterior is therefore equivalent to sampling the 1D distance posterior $g(r)$ and, for each distance sample, drawing one velocity sample from the 2D Gaussian $N_v(m_v, \Sigma_v)$. No approximation beyond the assumed Gaussian velocity prior is introduced.

The premise it rests on. The velocity prior is assumed to be a Gaussian distribution at each distance and direction; if the true velocity distribution of stars is significantly non-Gaussian, the conditional velocity posteriors would not be Gaussian and the closed-form 1D+2D reduction would no longer be exact.

If the paper is right (Pith's inferences, not the authors')

Kinegeometric distance and velocity estimates for Gaia DR3 (and future releases) can be computed 15 times faster than with the previous MCMC approach, with better accuracy and no convergence diagnostics.

The method eliminates the need for MCMC chain thinning, burn-in, and walker tuning for this problem, replacing them with direct 1D sampling via numerical inversion (e.g., PINV) and standard Gaussian random number generation.

The 1D+2D factorization makes it straightforward to compute marginal velocity posteriors as sums of Gaussians, enabling analytic or semi-analytic summaries of velocity distributions for large stellar samples.

Because the distance and velocity are often strongly correlated in the 3D posterior, the reduction to 1D+2D avoids the sampling difficulties that arise from that correlation in MCMC.

The method could be extended to non-Gaussian velocity priors by replacing the 2D Gaussian mixture with a more general distribution, but the closed-form Gaussian sum would be lost, requiring numerical integration over distance for each velocity sample.

For stars with very large parallax uncertainties (e.g., negative parallaxes), the 1D distance posterior $g(r)$ may be heavy-tailed; the PINV interpolation method used here may need careful handling of the tail to avoid bias in the velocity posterior.

The factor $Q(r,\varpi,\mu)$ (equation 9b) can be interpreted as a correction to the geometric posterior that incorporates proper-motion information; its shape as a function of $r$ determines how much the proper motions tighten the distance estimate relative to the parallax-only case.

If the Gaia astrometric solution is improved to reduce correlations between parallax and proper motion, the $X_\mu$ term in equation 5a becomes smaller, making the method even simpler (the velocity posterior mean becomes $k r \mu$ plus the velocity prior mean, weighted by inverse covariances).

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

When parallaxes are of poor quality, proper motions can improve stellar distance estimates using known constraints on stars' physical velocities. Formally this may be achieved via the kinegeometric posterior, a three-dimensional (3D) posterior in distance and velocity for given measured parallax and 2D proper motion. This posterior is the product of a covariant Gaussian likelihood in the parallax and proper motion with a distance and velocity prior that can be constructed from a Galaxy model. In earlier work, I showed how to infer this posterior for stars in Gaia DR3 by explicitly sampling it with a Markov Chain Monte Carlo method. Here I show that this posterior is equivalent to a 1D posterior in distance combined with a sum of 2D Gaussian distributions in velocity. This 1D+2D representation is both faster and more accurate to sample, thereby making the computation of the kinegeometric posterior for a large set of sources much easier.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-14

The authors' main claim, as the classifier picked it out 0.73

“This 1D+2D representation is both faster and more accurate to sample, thereby making the computation of the kinegeometric posterior for a large set of sources much easier.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.06 · states limitation 0.15 · human participants 0.02 · controlled comparison 0.06 · randomized assignment 0.01 · states quantitative result 0.05 · compares with prior results 0.87 · mentions code or data availability 0.03

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
When parallaxes are of poor quality, proper motions can improve stellar distance estimates using known constraints on stars' physical velocities. background 1.00 0.06 0.04 asserted 0.57
Formally this may be achieved via the kinegeometric posterior, a three-dimensional (3D) posterior in distance and velocity for given measured parallax and 2D proper motion. background 0.81 0.10 0.05 hedged possibility 0.95
This posterior is the product of a covariant Gaussian likelihood in the parallax and proper motion with a distance and velocity prior that can be constructed from a Galaxy model. background 0.66 0.13 0.03 asserted 0.98
In earlier work, I showed how to infer this posterior for stars in Gaia DR3 by explicitly sampling it with a Markov Chain Monte Carlo method. background 0.85 0.73 0.04 asserted 1.00
Here I show that this posterior is equivalent to a 1D posterior in distance combined with a sum of 2D Gaussian distributions in velocity. contribution 0.65 0.90 0.69 asserted 1.00
This 1D+2D representation is both faster and more accurate to sample, thereby making the computation of the kinegeometric posterior for a large set of sources much easier. interpretation 0.52 0.86 0.71 asserted 1.00

What kind of paper the classifier reports this is: method or tool 0.93 · evidence from proof or derivation 0.82

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-14: 1 request, 4,965 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (3)
Figure 1 of the paper
Figure 1. Finally, the conditional velocity posterior at a given distance r is P ∗ (v | r) = P ∗ (v, r) P∗ (r) = Nv(mv, Σv) ≡ f(v, r) (14) which comes from dividing equation 11a by equation 12b. This is a 2D Gaussian with mean and covariance from equation 10 evaluated at the specified distance. Step 6: Posterior summaries. From the previous steps we can produce a set of distance samples Ng {ri} and a cor￾responding set of Ng velocity samples {vi}. From these we may compute summary statistics, such as means and variances, 1D quantiles on the distance and velocities, as well as correlation
Figure 2 of the paper
Figure 2. Kinegeometric posteriors for a star with
Figure 3 of the paper
Figure 3. Comparison of various summaries (one per panel) of the distance and velocity posteriors for

Figures are the authors', reproduced from the paper.

References (6)
  1. [1] Bailer-Jones C.A.L., 2015.Estimating distances from parallaxes: a tutorial. PASP 127, 994 [paper I] Bailer-Jones C.A.L., Rybizki J., Fouesneau M., Mantelet G., Andrae R., 2018.Estimating distances from parallaxes. IV . Distances to 1.33 billion stars in Gaia data release on Pith
  2. [2] AJ 156, 58 [paper IV] Bailer-Jones C.A.L., Rybizki J., Fouesneau M., Demleitner M., Andrae R., 2021.Estimat- ing distances from parallaxes. V . Geometric and photogeometric distances to 1.47 billion stars in Gaia Early Data Release on Pith
  3. [3] AJ 161, 147 [paper V] Bailer-Jones C.A.L., 2023.Estimating distances from parallaxes. VI. A method for inferring distances and transverse velocities from parallaxes and proper motions demonstrated on Gaia Data Release on Pith
  4. [4] AJ 166, 269 [paper VI] Derflinger G., H ¨ormann W., Leydold J., 2010.Random variate generation by numerical inversion when only the density is known. ACM Transactions on Modeling and Computer Simulation (TOMACS), 20(4), 18 Foreman-Mackey D., Hogg D.W., Lang D., Goodman J., 201... on Pith
  5. [5] Summary of the contents and survey proper- ties. A&A 616, A1 Gaia Collaboration, 2021.Gaia Early Data Release on Pith
  6. [6] The astrometric solution. A&A 616, A2 Luri X., Brown A.G.A., Sarro L. M., Arenou F., Bailer-Jones C. A. L., et al. 2018.Gaia Data Release on Pith

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment