{"id":"512fd879-0d76-42d3-a793-a028c580dd02","arxiv_id":"1908.06602","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"A new stick-breaking prior class whose weight sequence is driven by a dependent Beta Markov chain, bridging the Dirichlet process and the Geometric process.","lead":"The authors introduce a new family of Bayesian nonparametric priors, the Beta-Binomial stick-breaking process, whose weights are generated by a dependent Markov chain. A single tuning parameter interpolates between two known priors, giving practitioners flexible control over how ordered the mixture weights are.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Gibbs V-update not implementable: Step 2's conditional is a Markov chain, not a product of Beta mixtures.","rationale":"The reader's weakest assumption is the full-support claim via Bissiri-Ongaro (2014). That is not the most load-bearing concern: the condition checked in Section 3 (positive probability that every length variable lies in an interval) is a known sufficient condition for full support in species sampling priors, and the check via conditioning on the Binomial chain is valid. The more serious issue is internal to the computational method. In Section 4.1, Step 2, the full conditional of V given the data and with U integrated out is a first-order Markov chain because the latent Binomial variables x_i enter through Bin(x_i|κ,v_i), which depends on v_i. The paper's assertion that this density 'can be normalized to a product of Beta densities mixtures' is false for κ≥1, as a simple κ=1 calculation shows. Without a valid sampling step for V, the claimed Gibbs sampler does not target the correct posterior, and the Section 5 illustrations are not reproducible. The theoretical core of the paper—Proposition 3.1 (weights sum to one), Theorem 3.2 (recovery of Dirichlet and Geometric processes), and the prior construction—appears sound, so the appropriate verdict remains conditional on correcting the sampler description and re-running the empirical demonstrations.","tokens_in":18753,"tokens_out":26246,"duration_ms":277619,"concrete_test":"Implement the Section 4.1 sampler exactly as described, with the V-update replaced by independent draws from the claimed 'product of Beta mixtures' (e.g., sampling each v_i marginally from its Beta mixture), and run it on the Section 5.1 dataset (database 1). Compare the resulting posterior density estimates and K_n distributions with those from the same sampler augmented by latent Binomial variables x_i (or using a forward-filtering/backward-sampling V-update), with identical truncation and MCMC length. If the estimates differ by more than Monte Carlo error, the published V-update is not targeting the true posterior.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Section 4.1, Step 2, the authors derive the marginal conditional of V with U integrated out: π(V|...) is proportional to Be(v1|α+α1, θ+θ1) times a product over i≥1 of sums over x of Be(v_{i+1}|α_{i+1}+α+x, θ_{i+1}+θ+κ−x) times constants times Bin(x|κ,v_i). They state that 'which can be normalized to a product of Beta densities mixtures'. This is incorrect: the factor Bin(x|κ,v_i) depends on v_i, coupling v_i and v_{i+1}. For example, with κ=1, α=θ=1 and no data, the transition density is 2[(1−v_1)(1−v_2)+v_1 v_2], which cannot be written as h(v_1)g(v_2); the cross term forces a Markov structure. Thus the joint law of V is a first-order Markov chain, not a product. Consequently, the described block step—sample V from this density, then U from U(0,w_{d_k})—is not implementable as written. A valid sampler would require either augmenting with the latent Binomial variables x_i (making V conditionally independent Betas given X) or using forward-filtering/backward-sampling. Neither is provided. Therefore the density-estimation algorithm in Section 5 is not reproducible from the paper, and the empirical support for BBSB mixtures cannot be assessed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a new class of discrete random probability measures, the Beta-Binomial stick-breaking (BBSB) priors, obtained by inserting a stationary Beta Markov chain into the stick-breaking construction. The Markov chain is built through a Beta-Binomial latent data augmentation following Pitt et al. (2002), with dependence parameter kappa. The authors prove conditional moments and correlation for the chain, show that kappa=0 yields independent Beta length variables (and with alpha=1 recovers the Dirichlet process), and that kappa goes to infinity yields the Geometric process as a weak limit. They prove that the stick-breaking weights sum to one almost surely, assert full support via Bissiri and Ongaro (2014), and develop a slice-sampling Gibbs algorithm for density estimation, including posterior inference for kappa. The algorithm is illustrated on simulated data. The theoretical proofs for the moments, limits, and weights summing to one are self-contained and appear correct.","tokens_in":18991,"tokens_out":16890,"duration_ms":163098,"significance":"If the construction is validated, BBSB priors provide a natural one-parameter family interpolating between Dirichlet and Geometric stick-breaking priors, with the dependence parameter controlling weight ordering and label-switching. This is a useful addition to the Bayesian nonparametrics toolbox, and the paper includes careful proofs of the main distributional properties and of the limiting cases. The key weakness is that the proposed MCMC algorithm is not correctly derived, so the empirical claims in Section 5 are not yet reproducible; this is a correctable issue rather than a flaw in the prior construction itself.","major_comments":[{"comment":"The conditional density stated for V is not a product of Beta mixtures. The displayed density for pi(V|...) is proportional to Be(v1|alpha+alpha_1,theta+theta_1) times product over i>=1 of sum_{x=0}^kappa Bin(x|kappa,v_i) Be(v_{i+1}|alpha_{i+1}+alpha+x,theta_{i+1}+theta+kappa-x) times constants. Because Bin(x|kappa,v_i) depends on v_i, the factor for index i couples v_i and v_{i+1}; the joint law is a first-order Markov chain, not a product over i. For example, with kappa=1, alpha=theta=1 and no data, the joint density of (v_1,...,v_m) is proportional to prod_{i=1}^{m-1} [(1-v_i)Be(v_{i+1}|1,2)+v_i Be(v_{i+1}|2,1)], which cannot be factorized as prod h_i(v_i). Consequently the block step 'sample V from this density, then U from U(0,w_dk)' is not implementable as written, and the density-estimation algorithm in Section 5 is not reproducible from the paper. A valid sampler would need to augment with the latent Binomial variables x_i (making V conditionally independent Betas given X and D) or use forward-filtering/backward-sampling; neither is supplied.","section":"Section 4.1, Step 2"},{"comment":"The full-support claim is delegated to Bissiri and Ongaro (2014) with the sentence 'the above observation shows that any Beta-Binomial prior has full support.' The paper displays only a finite-window positive-probability condition, P(cap_{i=1}^n (delta0 for every n and 0=1, but the sampler only stores V up to the truncation phi from Remark 4.1. The paper does not explain how to compute or approximate this infinite product, nor how to account for the unsampled tail variables v_i for i>phi. This is needed for the posterior inference for kappa reported in Section 5.2, whose results depend on the update in Eq. (16). Please specify a truncation scheme or an exact marginalization of the tail.","section":"Section 4.3, Eq. (16)"}],"minor_comments":[{"comment":"The Geometric process is described as a particular case of the BBSB prior, but Theorem 3.2(ii) only establishes a weak limit as kappa goes to infinity; for no finite kappa is the Geometric process obtained. Please adjust the wording to say 'limiting case'.","section":"Abstract and Introduction"},{"comment":"The phrase 'two of reversible Markov chains' appears to be a typo; it should read 'two reversible Markov chains'.","section":"Section 2, first paragraph"},{"comment":"The phrase 'which can be normalized to a product of Beta densities mixtures' is the incorrect claim addressed in Major Comment 1; in addition, 'latter sample U' should read 'later sample U'.","section":"Section 4.1, Step 2"},{"comment":"Lemma A.4 proves L2 convergence of x_n/n, but Proposition 2.2(ii) uses an almost-sure convergent construction. Please clarify that this follows by taking a subsequence or by a Skorokhod representation argument, since the proof as written only gives L2 convergence.","section":"Appendix A.4, proof of Lemma A.4"}],"recommendation":"major_revision","confidential_remarks":"The theoretical construction and the proofs of the main distributional properties are sound; the risk is concentrated in the MCMC development. The incorrect Step 2 of Section 4.1 can be fixed by augmenting with the latent Binomial variables, and the full-support claim can be completed by stating the cited theorem. If the authors address these points, the paper could be suitable. I would encourage the editor to invite a revision rather than reject."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: the paper has a genuinely new construction and the theoretical proofs are mostly solid, but the computational centerpiece has a real error that makes the proposed Gibbs sampler unimplementable as written. Worth a referee's time, but heavy revision needed.\n\nWhat's new: the Beta-Binomial stick-breaking prior, where stick-breaking lengths come from a stationary Beta Markov chain with Binomial latent variables. That's not in the cited literature, and it's not a reparameterization of an existing prior. The Dirichlet process (κ=0, α=1) and Geometric process (κ→∞) arise as genuine boundary/limit cases, proven in Theorem 3.2. I checked the appendix proofs of Proposition 2.1 (moments/correlation), Proposition 3.1 (weights sum to one), and the convergence result—they're clean. The intuition about κ controlling weight ordering is well supported by the simulation figures.\n\nSoft spots: the full-support claim is delegated to Bissiri and Ongaro (2014) with only a finite-window positive-probability condition shown. The authors don't verify that the cited theorem covers dependent, non-exchangeable length-variable sequences. That's a gap, but probably fixable if the theorem's hypotheses are broad enough.\n\nThe bigger problem is in Section 4.1, Step 2. The authors state that after marginalizing U, the conditional density of V 'can be normalized to a product of Beta densities mixtures.' That's wrong. The Bin(x|κ, v_i) factors in the product over i couple v_i and v_{i+1}. For κ=1, α=θ=1, no data, the joint density of (v_1,v_2) is proportional to (1−v_1)(1−v_2)+v_1v_2, which is not a product of functions of the individual variables. So the joint law is a first-order Markov chain, not independent components. The block Gibbs step as written cannot be sampled. A valid sampler would need either the latent Binomial variables augmented in (making V conditionally independent Betas given X) or a forward-filtering/backward-sampling step. Neither is given. Since the entire empirical section runs this sampler, those results are not reproducible from the paper.\n\nThe empirical section is also thin—simulated data only, no code, no exact generation settings, one comparison with Pitman-Yor. But that's a minor issue next to the sampler.\n\nBottom line: the theoretical construction is novel and the core limit results are solid. The paper deserves peer review, but a referee should require the sampler to be fixed or replaced, and the empirical claims rerun. I'd want to see that before citing the algorithm.","headline":"Novel BBSB prior with solid theory, but the Gibbs sampler as written is not implementable and the empirical claims need a rerun.","tokens_in":19547,"tokens_out":3527,"would_cite":true,"duration_ms":31100,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["62G07","60G57","62F15"],"pacs":[],"model":"deepseek-v4-flash","headline":"A single dependence parameter in a Beta-Binomial stick-breaking prior interpolates between the Dirichlet process and the Geometric process prior.","keywords":["Beta-Binomial Markov chain","stick-breaking prior","Dirichlet process","Geometric process","density estimation","Bayesian nonparametrics","species sampling prior","MCMC"],"falsifier":"Compute whether a BBSB prior with, say, $\\kappa=1$, $\\alpha=\\theta=1$ assigns positive probability to every weak neighbourhood of a fixed target measure such as $0.5\\delta_0+0.5\\delta_1$; if any target measure has zero prior neighbourhood probability, the full-support claim fails as stated.","tokens_in":18537,"feed_emoji":"📊","tokens_out":10329,"duration_ms":92832,"temperature":0.7,"pith_summary":"This paper proposes a new family of Bayesian nonparametric priors, the Beta-Binomial stick-breaking (BBSB) prior, by making the length variables in a stick-breaking construction dependent rather than independent. The length variables come from a stationary Beta Markov chain whose dependence parameter $\\kappa$ controls how similar consecutive sticks are. The paper's central claim is that tuning $\\kappa$ moves the prior between two well-known extremes: $\\kappa=0$ with $\\alpha=1$ recovers the Dirichlet process, and as $\\kappa\\to\\infty$ the prior converges to the Geometric process prior. If true, this gives density-estimation and clustering users a single knob that balances efficient component mixing against ordered, identifiable weights, and the paper provides a Gibbs sampler to use the prior in practice.","feed_headline":"One parameter tunes Bayesian priors from Dirichlet to Geometric","feed_subtitle":"Tuning stick-length dependence balances mixture mixing with weight identifiability.","key_machinery":"The central object is the Beta-Binomial Markov chain: a stationary first-order Markov chain $V=(v_i)_{i\\ge1}$ with Beta($\\alpha,\\theta$) marginals, built from the Beta-Binomial conjugate pair by setting $x_i\\mid v_i\\sim\\mathrm{Bin}(\\kappa,v_i)$ and $v_{i+1}\\mid x_i\\sim\\mathrm{Be}(\\alpha+x_i,\\theta+\\kappa-x_i)$. The integer $\\kappa$ is the dependence parameter; Proposition 2.1 gives the correlation $\\rho_{v_i,v_{i+1}}=\\kappa/(\\alpha+\\theta+\\kappa)$, so $\\kappa=0$ yields independent lengths and large $\\kappa$ yields nearly identical lengths. Inserting these dependent lengths into the stick-breaking formula produces the BBSB weights, and the chain is what carries the argument: its dependence structure is exactly the mechanism that interpolates between the Dirichlet and Geometric process priors and controls whether the weights appear in size-biased or decreasing order.","core_discovery":"The paper's central claim is that a new class of random probability measures, the Beta-Binomial stick-breaking priors, arises when the stick-breaking weights $w_j=v_j\\prod_{i=1}^{j-1}(1-v_i)$ are built from length variables $v_i$ that follow a strictly stationary Beta($\\alpha,\\theta$) Markov chain rather than being independent. The chain is generated through a latent Binomial chain, and its integer parameter $\\kappa$ is the dependence knob: consecutive lengths have correlation $\\kappa/(\\alpha+\\theta+\\kappa)$, so $\\kappa=0$ makes them independent. With $\\alpha=1$, that extreme reproduces the Dirichlet process, while as $\\kappa\\to\\infty$ the lengths converge to a single Beta($\\alpha,\\theta$) value and the prior converges in distribution to the Geometric process prior. For every $\\kappa$, the weights sum to one almost surely, and the prior is claimed to have full support via a known result for species sampling priors. The dependence parameter also controls weight ordering: the Dirichlet extreme gives size-biased weights, the Geometric extreme gives decreasingly ordered weights, and intermediate values offer a practical trade-off.","pith_inferences":["Because the Beta marginals are fixed while only the dependence changes, the same construction could generate dependent families of stick-breaking priors for time-series or non-exchangeable data, where the ordering of clusters itself is informative.","The paper proves $P[w_{j+1}