Pith. sign in

REVIEW 3 major objections 5 minor 31 references

Gaia Parallax of Milky Way Globular Clusters -- A Solution of Mixture Model

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read A two-component mixture model recovers the mean parallax of 120 globular clusters from Gaia DR2 parallaxes alone, sidestepping cluster membership determination entirely.

desk verdict Useful membership-free GC parallax catalog, but the field-profile assumption needs stronger validation. read the letter →

arxiv 1908.06031 v1 pith:RP3EDNSO submitted 2019-08-16 astro-ph.GA astro-ph.IM

classification astro-ph.GAastro-ph.IM
keywords parallaxesglobularclusters:generalGaiaDR2mixturemodeldirectparallaxzero-pointdistancedeterminationnestedsampling
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

This paper tries to establish that the mean parallax of a globular cluster can be recovered from Gaia DR2 parallax measurements alone, without deciding which stars belong to the cluster. The authors build a two-component mixture model that treats cluster stars as a sharp spike at one parallax and field stars as a broad asymmetric background, then fit it to the parallaxes of all stars in the cluster's projected area. Validated on simulated clusters and applied to real data, the method resolves 120 of the 147 globular clusters in the input catalog, more than 80% of the identified Milky Way globular clusters. If the method is sound, it offers an independent, membership-free route to cluster distances and a new way to check the Gaia parallax zero-point.

What carries the argument

The central object is the two-component parallax mixture model. The cluster component is a delta function at the mean parallax $\varpi_c$; the field component follows the exponentially decreasing density profile of Bailer-Jones et al. (2018), $\varphi_f(\varpi)=\frac{\varpi_L^3}{2(\varpi-\varpi_0)^4}\exp\left(-\frac{\varpi_L}{\varpi-\varpi_0}\right)$ for $\varpi>\varpi_0$. Each component is convolved with the per-star Gaussian parallax error, and the model is fitted by nested sampling, with $\varpi_c$ obtained by marginalizing over $\varpi_0$, $\varpi_L$, and the cluster fraction $n_c$. The asymmetry of the field profile is what lets the model peel a symmetric cluster signal out of a heavily contaminated, error-dominated parallax histogram.

What would settle it

Fit the mixture model to synthetic sight lines drawn from a realistic Galactic model whose field-star parallax distribution is not exponential (e.g., including bulge and spiral-arm structure) over the same distance and contamination ranges as the Pancino et al. (2017) simulations; if the recovered $\varpi_c$ deviates from the input by more than the 1-5% claimed accuracy, the exponential field profile is the cause. Alternatively, test a real line of sight where a known foreground structure creates a secondary bump in the parallax histogram and check whether the mixture model misattributes it to the cluster.

Watch

Extended reading notes

Core claim

The central claim is that a mixture model fitted to raw parallax histograms alone can separate a cluster's mean parallax $\varpi_c$ from field-star contamination without membership determination. The cluster is modeled as a delta function at $\varpi_c$, the field by the skewed exponentially decreasing density profile of Bailer-Jones et al. (2018), and each is convolved with per-star Gaussian errors. Nested sampling fits the four parameters and then marginalizes out the field parameters to obtain a posterior for $\varpi_c$. On the 18 simulated clusters of Pancino et al. (2017) the recovered $\varpi_c$ matches the input and remains stable when the fitted aperture is varied. Applied to Gaia DR2, 120 globular clusters are resolved, and comparison with indirect distances gives an offset of $-27.6\pm1.7\,\mu\mathrm{as}$ and a scatter of $22.8\pm1.3\,\mu\mathrm{as}$, consistent with the known Gaia DR2 zero-point and its variation.

Load-bearing premise

The separation of cluster from field rests on the assumed analytic form of the field-star parallax distribution (Eq. 3, a simple exponentially decreasing density profile); if the true field in a cluster direction is more complex, the fitted cluster parallax can absorb the model mismatch and become biased.

Editorial extensions

If this is right

  • The mixture model can be applied to future Gaia data releases with no membership determination, so the direct-parallax sample of globular clusters will grow as astrometric precision improves.
  • The 120 resolved clusters constitute the largest set of independent direct parallaxes, providing tracers for monitoring the Gaia zero-point and its spatial variation.
  • For the nearest clusters such as M 4 and NGC 6397, the statistical precision reaches a fraction of a percent, comparable to or better than many indirect distance methods.
  • In heavily contaminated fields like M 62, the mixture model avoids the bias that membership-based direct measurements can suffer, because it does not need to discard field stars.
  • The measured offset of about $-27.6\,\mu\mathrm{as}$ between apparent mixture-model parallaxes and indirect distances independently confirms the global Gaia DR2 zero-point.

Reading between the lines

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

  • If the field-star model were replaced by a more realistic multi-component Galactic model (disc, bulge, halo), the method could likely resolve more of the 27 rejected clusters and reduce the scatter in the zero-point comparison.
  • The same mixture approach could be applied to open clusters, dwarf spheroidal galaxies, or any compact stellar system whose mean parallax is contaminated by foreground and background stars in Gaia-like catalogs.
  • Because the authors caution that the fitted $\varpi_0$ is only nominal, a natural extension is to calibrate the field profile against quasar-based zero-point maps and turn the apparent cluster parallaxes into absolute distances with per-cluster corrections.
  • The significant disagreements with H18 for NGC 6397, $\omega$ Cen, and M 62 indicate method-dependent biases; a combined analysis using colour-magnitude-based membership posteriors would show whether the mixture model or the membership approach is wrong in those fields.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 proposes a two-component mixture model to derive the mean parallax of Milky Way globular clusters directly from Gaia DR2 parallaxes, thereby avoiding explicit membership determination: the cluster term is a delta function convolved with the per-star parallax errors, and the field term follows the exponentially decreasing density profile of Bailer-Jones et al. (2018), Eq. (3). The model is validated on the 18 simulated clusters of Pancino et al. (2017), applied to 147 clusters from Kharchenko et al. (2013), and 120 are reported as successfully resolved. The fitted apparent parallaxes are compared with the direct parallaxes of Gaia Collaboration et al. (2018) and with indirect distance catalogs; a combined offset of -27.6 +/- 1.7 microarcseconds and scatter of 22.8 +/- 1.3 microarcseconds are interpreted as a confirmation of the known Gaia DR2 parallax zero-point and its variation. The paper also discusses distances of five nearby clusters after applying a -29 microarcsecond zero-point correction.

Significance. If the central claim is correct, this is a genuinely useful contribution: it provides the largest direct-parallax catalog of Milky Way globular clusters and offers an independent route that does not depend on proper-motion or CMD membership selection. The likelihood, nested-sampling implementation, marginalization over nuisance parameters, and validation on the P17 simulated clusters are clearly described, and the authors are honest that the fitted zero-point parameter is only nominal and cannot be used directly for the Gaia zero-point. The comparison with independent catalogs is appropriate for a first demonstration, and the claimed offset is consistent with the known Gaia DR2 zero-point. The main limitation is that the robustness of the recovered cluster parallax against misspecification of the field-star parallax distribution is not established for real Milky Way fields, which is exactly the regime where the method is intended to be used.

major comments (3)
  1. [Sec. 2.2, Eq. (3)] The load-bearing assumption is that any mismatch between the true field-star parallax distribution and the simple exponentially decreasing profile of Eq. (3) is absorbed by the nuisance parameters n_c, omega_L, and omega_0, rather than leaking into the cluster parallax omega_c. The paper itself notes that omega_0 shifts to compensate for model mismatch, but the claim that omega_c is unaffected is demonstrated only on the 18 P17 simulated clusters, each of which has a single background component (halo, disc, or bulge). Real fields, especially toward the bulge, can contain multiple overlapping stellar populations and extinction structure, and the cluster term - a symmetric Gaussian whose width is tied to the per-star sigma_pi - is the only remaining flexible symmetric component that can absorb an asymmetric field residual. I recommend adding a quantitative test in which synthetic cluster populations with known parallax and known n_c are injected into real Gaia DR2 field-star samples (e.g., from annuli around clusters or from adjacent control fields) and the recovery of omega_c is checked as a function of field complexity. Without such a test, the validity of the 120-cluster catalog and of the -27.6 +/- 1.7 microarcsecond zero-point offset is not fully established for realistic fields.
  2. [Sec. 3, radius selection and cluster rejection] The construction of the final catalog involves data-dependent selection choices that are not covered by the P17 simulation validation. The fitting radius is chosen from r1, (r0+r1)/2, or r0 based on the overdensity and the stability of the fits, and 27 of 147 clusters are rejected because the marginal PDF of omega_c is not single-mode or because the result is not stable over a range of radii. Because both the radius choice and the acceptance criterion are made after inspecting the same data, there is a risk of selection bias: clusters with problematic fields may be preferentially excluded, while the reported scatter and offset are computed on the accepted sample. I ask the authors to report, for every cluster, which radius was used and why, to quote the number of clusters that failed at each selection stage, and to show for all 120 accepted clusters that varying the radius within the stated range changes omega_c by less than the statistical error. It would also be useful to test whether acceptance or rejection correlates with position on the sky, distance, or field density, since such a correlation would indicate that the field model, not the cluster signal, is driving the selection.
  3. [Sec. 4.1, comparison with H18] The comparison with the H18 membership-based parallaxes shows that 9 of 75 clusters (12%) exceed the 2-sigma combined uncertainty and 5 exceed 3-sigma. The authors attribute part of this to contamination in the H18 membership sample and discuss M 62 as a specific case. However, the opposite failure mode is never tested: a cluster component created by unmodeled field structure could produce a spurious omega_c, and because the method uses no membership information, this possibility cannot be excluded a priori. The M 62 discussion shows that membership contamination can be severe, but it does not establish that the mixture model cannot produce a false positive in dense bulge fields. I recommend testing for false positives by applying the mixture model to control fields with no known cluster, or by injecting synthetic clusters of known parallax and measuring the rate of spurious detection as a function of field density. Without this test, the 12% discrepant fraction is more naturally read as evidence that at least one of the two methods is biased in a nontrivial subset of fields, and the direction of the bias is not determined by the comparison presented.
minor comments (5)
  1. [Sec. 2.2, Eq. (8)] The text refers to the 'joint logistical likelihood'; this should read 'joint log-likelihood' or 'logarithmic likelihood'.
  2. [Sec. 2.2 and Fig. 4] The sentence 'Fitting results are plotted in figure 1' appears to refer to the 18-simulated-cluster comparison shown in Figure 4; the cross-reference should be corrected.
  3. [Sec. 3] The outlier cut 'omega + 0.029 mas < -3 sigma_pi' is dimensionally confusing as written; please restate with explicit units, e.g., 'omega/arcsec + 0.000029 < -3 sigma_pi'.
  4. [Sec. 4.2] When converting H10 distance moduli to parallaxes, the paper states that a 0.1 mag error is assumed for all clusters, but it does not describe how this error is propagated to the parallax comparisons; please specify the conversion and the treatment of asymmetric distance errors.
  5. [Abstract] The phrase 'They construct the largest direct parallax sample up to now' is awkward because the antecedent is unclear; consider 'This sample constitutes the largest direct-parallax catalog of globular clusters to date.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the cluster parallax is a fitted parameter of an explicit mixture model, validated on external simulations and compared with independent measurements after the fit.

full rationale

The paper's central output, the cluster parallax ϖc, is an explicit free parameter of a likelihood built from an assumed mixture of a delta-function cluster component (Eq. 2) and a field component (Eq. 3), with individual errors convolved as Gaussians (Eqs. 4–8). The value is estimated by Nested Sampling from the Gaia parallax data themselves; it is not defined in terms of, or constrained by, the indirect distances, the H18 results, or the Gaia zero-point discussed later. Validation is external: 18 simulated clusters of Pancino et al. (2017) with known input parallaxes are recovered, and the recovered values are compared with the true inputs, not with the model outputs. The comparisons with H18 direct parallaxes and with indirect distances (H10, post-H10) are performed after fitting, and the resulting offset (−27.6 ± 1.7 μas) is reported as a measurement consistent with the known Gaia zero-point; that zero-point is not inserted into the model to force the result. There are no load-bearing self-citations, no uniqueness theorem imported from the authors' prior work, and no fitted quantity is renamed as a prediction. The acknowledged modeling limitation—that Eq. (3) is an approximation and the fitted ϖ0 is only nominal—raises a genuine identifiability risk that field-model misspecification could bias ϖc, but that is a statistical robustness concern explicitly flagged by the authors, not a circular reduction of the derivation to its inputs. The derivation chain is therefore self-contained and non-circular.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central result depends on a small set of fitted nuisance parameters and on the assumed analytical form of the field-star parallax distribution. The paper does not introduce new physical entities; its contribution is a statistical measurement method. The main extras beyond standard Bayesian fitting are the domain assumptions about the field profile and about the effect of incompleteness.

free parameters (5)
  • Cluster parallax omega_c = fitted per cluster; example 499.5 microarcseconds for M4
    The central fitted parameter of the mixture model; the paper's final reported value.
  • Cluster fraction n_c = fitted per cluster
    Mixture weight of the cluster component; nuisance parameter marginalized over in the quoted omega_c.
  • Field scale parameter omega_L = fitted per cluster
    Inverse scale length in Eq. (3); fitted nuisance parameter controlling the width of the field-star parallax distribution.
  • Field zero-point parameter omega_0 = fitted per cluster
    Lower-limit parameter in Eq. (3); explicitly labeled as only a nominal zero-point that cannot be used to assess the Gaia zero-point directly.
  • Fitting radius choice = chosen per cluster from r1, (r0+r1)/2, r0
    The aperture is selected based on stability of the fit, a data-dependent choice that could affect the sample.
assumptions (6)
  • domain assumption The intrinsic cluster parallax distribution is a delta function (Eq. 2).
    Cluster size is taken as negligible compared with its distance; this is standard for GCs but ignores any internal spread.
  • domain assumption Field-star parallaxes follow the exponentially decreasing density profile of Eq. (3) (Bailer-Jones et al. 2018).
    This approximation drives the asymmetry used to separate field from cluster; authors acknowledge it is not exact.
  • domain assumption Gaia DR2 parallax uncertainties are Gaussian and known after scaling by 1.081 (Lindegren et al. 2018).
    The likelihood in Eq. (7) uses the reported sigma_omega as the Gaussian width; the scaling factor is adopted from prior calibration.
  • standard math The joint likelihood is a product of independent mixture densities and Nested Sampling yields valid marginal posteriors.
    Bayesian machinery; the paper does not specify the prior distributions used for the four parameters.
  • domain assumption Radial and magnitude-dependent incompleteness changes only the cluster fraction n_c, not the fitted omega_c.
    Stated without proof in Section 3; important for crowded clusters such as omega Cen and 47 Tuc.
  • domain assumption The cluster centers and characteristic radii from Kharchenko et al. (2013) are accurate.
    The input catalog defines the apertures and target list of 147 GCs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Gaia Parallax of Milky Way Globular Clusters -- A Solution of Mixture Model." pith.science (2026). https://pith.science/paper/RP3EDNSO

@misc{pith2026190806031,
  author       = {Pith},
  title        = {Pith review of: Gaia Parallax of Milky Way Globular Clusters -- A Solution of Mixture Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RP3EDNSO}},
  note         = {Machine review of arXiv:1908.06031}
}
abstract

We have established a mixture model approach to derive the parallax of the Milky Way globular clusters. It avoids the problem of cluster membership determination and provides a completely independent astrometrical solution by purely using the parallax data. This method is validated with simulated clusters of \cite{2017MNRAS.467..412P}. We have resolved 120 real globular clusters by the mixture model using parallaxes of the second data release of \gaia. They construct the largest direct parallax sample up to now. In comparing with other direct parallax results based on cluster members, including 75 clusters of \cite{2018A&A...616A..12G}, our method presents its accuracy, especially for some particular clusters. A systematic offset of $-27.6\pm1.7$ $\mu$as, together with a scatter of $22.8\pm1.3$ $\mu$as is found in comparing with other indirect parallax measurements. They are consistent with the global value and the variation of the zero-point of current \gaia parallaxes. Distances of several specific nearby globular clusters are discussed while the closest ones can reach high precisions, even taking the systematic error into account.

Figures

Figures reproduced from arXiv: 1908.06031 by the authors.

Figure 1
Figure 1. Model distribution of parallaxes. The top and middle panels are for the cluster members and the field stars separately. The solid lines show the intrinsic distributions of φc and φf . The dashed, dash-dotted and dotted lines show the apparent distributions of ϕc and ϕf with observational errors of σ$ = 300, 600, 900 µas respectively. The bottom panel shows the com￾binations of the cluster members (red line) or the f… view at source ↗
Figure 2
Figure 2. Probability density functions (PDFs) of parameters based on the Nested Sampling approach in the fitting of a simulated GC at d = 15kpc with contamination of disc field stars (GC 06 of P17). Sample stars are selected within 3rh. White cross symbols indicate the maximum likelihood point. Histograms show the 1-dimensional marginalized PDFs for each parameter, with vertical grey lines and shadows represent the mean valu… view at source ↗
Figure 3
Figure 3. Histograms of parallax distributions of the simulated cluster of GC 06 of P17. The cluster members, the field stars and their mixture are shown as dark grey line, grey shadow and light grey shadow separately. Corresponding apparent model distributions of these two components and their mixture are also shown for comparison. where N($0 ; $, σ$) represents the Gaussian probability centred at $ and the subscript ’x’ inf… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison of 18 simulated GCs of P17. ∆$c = $c,fitting − $c,true is the difference between the fitting result and the true input value. GCs at 5, 10 and 15 kpc are lined from top to bottom, and different contaminations of the halo, disc and bulge, are arranged from le…
Figure 5
Figure 5. Figure 5: Histograms of parallax distributions of selected GCs. Model distributions of the cluster (red), the field (blue) and their mixture (black) are shown for comparison. Panels (a)-(g) are GCs at different distances, with which can show the various of the relationships of t…
Figure 6
Figure 6. Figure 6: Comparison of the mean parallax of GCs between this work (TW) and other direct parallax results. ∆$c = $c,TW −$c,others, and the error bars σ$c,syn represent the quadratic sum of both of their uncertainties. Circles are 75 GCs from H18, with grey filled for |∆$c| > 2σ$…
Figure 7
Figure 7. Figure 7: Comparisons of the parallaxes of GCs between this work (TW) and other indirect parallax results. Left panel: the comparison of H10. A typical error of 0.10 mag of the distance modulus is adopted for all H10 values. Hollow circles are the 75 common GCs of H18, and grey …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 27 canonical work pages

  1. [1]

    A., Bustos Fierro I

    Arellano Ferro A., Ahumada J. A., Bustos Fierro I. H., Calder \'o n J. H., Morrell N. I., 2018, AN, 339, 183

  2. [2]

    Arenou F., et al., 2018, A&A, 616, A17

  3. [3]

    Bailer-Jones C. A. L., Rybizki J., Fouesneau M., Mantelet G., Andrae R., 2018, AJ, 156, 58

  4. [4]

    Baumgardt H., Hilker M., 2018, MNRAS, 478, 1520

  5. [5]

    F., et al., 2015, ApJ, 799, 165

    Braga V. F., et al., 2015, ApJ, 799, 165

  6. [6]

    M., Casertano S., Strader J., Riess A., VandenBerg D

    Brown T. M., Casertano S., Strader J., Riess A., VandenBerg D. A., Soderblom D. R., Kalirai J., Salinas R., 2018, ApJ, 856, L6

  7. [7]

    G., Recio-Blanco A., Lucatello S., D'Orazi V., Cassisi S., 2010, A&A, 516, A55

    Carretta E., Bragaglia A., Gratton R. G., Recio-Blanco A., Lucatello S., D'Orazi V., Cassisi S., 2010, A&A, 516, A55

  8. [8]

    Chen S., Richer H., Caiazzo I., Heyl J., 2018, ApJ, 867, 132

Show all 31 references
  1. [9]

    A., Pritzl B

    Contreras R., Catelan M., Smith H. A., Pritzl B. J., Borissova J., Kuehn C. A., 2010, AJ, 140, 1766

  2. [10]

    P., Cameron E., Pettitt A

    Feroz F., Hobson M. P., Cameron E., Pettitt A. N., 2013, arXiv:1306.2144

  3. [11]

    Gaia Collaboration, et al., 2018, A&A, 616, A1

  4. [12]

    Gaia Collaboration, et al., 2018, A&A, 616, A12 (H18)

  5. [13]

    Gaia Collaboration, et al., 2016, A&A, 595, A1

  6. [14]

    E., 1996, AJ, 112, 1487 (H10)

    Harris W. E., 1996, AJ, 112, 1487 (H10)

  7. [15]

    Hernitschek N., et al., 2019, ApJ, 871, 49

  8. [16]

    Kaluzny J., et al., 2013, AJ, 145, 43

  9. [17]

    V., Piskunov A

    Kharchenko N. V., Piskunov A. E., Schilbach E., R \"o ser S., Scholz R.-D., 2012, A&A, 543, A156

  10. [18]

    V., Piskunov A

    Kharchenko N. V., Piskunov A. E., Schilbach E., R \"o ser S., Scholz R.-D., 2013, A&A, 558, A53

  11. [19]

    A., Feigelson E

    Kuhn M. A., Feigelson E. D., 2017, arXiv:1711.11101

  12. [20]

    L., Walker A., 2014, ApJS, 210, 6

    Lee J.-W., L \'o pez-Morales M., Hong K., Kang Y.-W., Pohl B. L., Walker A., 2014, ApJS, 210, 6

  13. [21]

    Lindegren L., et al., 2018, A&A, 616, A2

  14. [22]

    R., et al., 2015, ApJ, 808, 11

    Neeley J. R., et al., 2015, ApJ, 808, 11

  15. [23]

    M., Gilligan C., Chaboyer B., 2017, ApJ, 838, 162

    O'Malley E. M., Gilligan C., Chaboyer B., 2017, ApJ, 838, 162

  16. [24]

    Pancino E., Bellazzini M., Giuffrida G., Marinoni S., 2017, MNRAS, 467, 412 (P17)

  17. [25]

    Randich S., et al., 2018, A&A, 612, A99

  18. [26]

    F., Cudworth K

    Rees R. F., Cudworth K. M., 2017, AAS, 229, 343.07

  19. [27]

    A., Denissenkov P

    VandenBerg D. A., Denissenkov P. A., 2018, ApJ, 862, 72

  20. [28]

    A., Denissenkov P

    VandenBerg D. A., Denissenkov P. A., Catelan M., 2016, ApJ, 827, 2

  21. [29]

    L., van der Marel R

    Watkins L. L., van der Marel R. P., 2017, ApJ, 839, 89

  22. [30]

    L., van der Marel R

    Watkins L. L., van der Marel R. P., Bellini A., Anderson J., 2015, ApJ, 812, 149

  23. [31]

    A., et al., 2012, AJ, 143, 50

    Woodley K. A., et al., 2012, AJ, 143, 50

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.