Location-Dependent Stick-Breaking Prior

• Location-dependent stick-breaking priors are random probability measures that adjust mixture weights based on external indices, enabling predictor-specific clustering.
• The methodology employs variants like logit transformations, spatial-temporal kernels, and tree-based binary regressions to adapt weights dynamically.
• Applications in density regression, VAR models, and spatial prediction demonstrate enhanced local adaptation and computational efficiency compared to traditional methods.

A location-dependent stick-breaking prior is a family of random probability measures indexed by a predictor, location, or space-time coordinate, typically written $G_x=\sum_h w_h(x)\delta_{\theta_h}$, in which the atoms are shared across index values while the stick-breaking weights vary with $x$. In this usage, “location” may mean a baseline covariate vector $\bm z$, a spatial coordinate $s$, a spatio-temporal index $(s,t)$, or a more general input $z$; the defining feature is that the mixing distribution changes with the index through $w_h(x)$, not that the likelihood alone depends on covariates (Beraha et al., 2022, Grazian, 2023, Rigon et al., 2017). This places location-dependent stick-breaking within the broader class of dependent random probability measures, and, in the common-atoms form, it yields predictor- or location-dependent clustering through index-specific mixture probabilities rather than through index-specific atoms alone (Horiguchi et al., 2022, Linderman et al., 2015).

1. Definition and formal structure

In the ordinary Sethuraman representation of a Dirichlet process,

$$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$

the weights do not depend on any covariate or index value. Every observation sees the same $\{\pi_k\}$. A location-dependent stick-breaking prior replaces this exchangeable construction by an indexed family

$$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$

so that the random measure changes with $x$0 through the break proportions or their transforms (Grazian, 2023, Rigon et al., 2017).

The common-atoms form is especially prominent. In the child-growth VAR model,

$x$1

with global atoms $x$2 shared across subjects and covariate-dependent weights $x$3 (Beraha et al., 2022). In the spatio-temporal formulation,

$x$4

the same countable atoms are used for all $x$5, while the weight vector varies with location and time (Grazian, 2023). The same principle appears in covariate-dependent treeSB models,

$x$6

where each leaf weight is a path probability determined by covariate-dependent binary gates (Horiguchi et al., 2022).

This construction is distinct from models in which covariates affect only the kernel or only the atom locations. The sources repeatedly separate three mechanisms: common weights with varying atoms, varying weights with common atoms, and hybrids in which both vary. In location-dependent stick-breaking priors, the defining dependence is in the weights. This is why they are often described as predictor-dependent mixtures, spatial stick-breaking priors, or spatio-temporal stick-breaking processes, depending on the index set (Beraha et al., 2022, Grazian, 2023).

A useful latent-allocation view introduces $x$7 or $x$8 such that

$x$9

followed by a component-specific draw. This makes explicit that the prior acts on the random partition by changing allocation probabilities as the location index changes (Beraha et al., 2022, Rigon et al., 2017).

2. Principal constructions

Several non-equivalent constructions instantiate the same general idea: keep the stick-breaking normalization but let the breaks depend on predictors, locations, or paths in a tree.

Construction	Index	Dependence mechanism
Logit stick-breaking prior	$\bm z$0, $\bm z$1	$\bm z$2
Spatial or spatio-temporal stick-breaking	$\bm z$3, $\bm z$4	$\bm z$5
Tree stick-breaking	$\bm z$6	Node-specific binary regressions along tree paths
Finite multinomial logistic stick-breaking	$\bm z$7	$\bm z$8

The logit stick-breaking prior is the most direct regression-based version. In the LSBP density-regression model,

$\bm z$9

with $s$0 (Rigon et al., 2017). The same mechanism appears in finite form in the obesity application,

$s$1

which yields a covariate-dependent finite stick-breaking prior for subject-specific VAR coefficients (Beraha et al., 2022).

Spatial and spatio-temporal models instead modulate the breaks through kernels. In the spatial case,

$s$2

and in the spatio-temporal extension,

$s$3

with $s$4, $s$5, and $s$6 (Grazian, 2023). The kernel may be separable,

$s$7

or non-separable, as in the Gneiting-type example

$s$8

with $s$9 corresponding to separability (Grazian, 2023).

Tree-based generalizations reinterpret stick-breaking as a bifurcating routing architecture. For a binary string $(s,t)$0,

$(s,t)$1

and, in the covariate-dependent form,

$(s,t)$2

(Horiguchi et al., 2022). This formulation subsumes the standard sequential construction as a lopsided tree and introduces balanced trees as alternative topologies.

A related but finite-$(s,t)$3 construction appears in dependent multinomial models: $(s,t)$4 with latent Gaussian structure on $(s,t)$5 over documents, time, space, or covariates (Linderman et al., 2015). Although this is not a BNP prior over infinitely many atoms, it is a genuine location-dependent stick-breaking model on the simplex.

3. How location dependence changes clustering and dependence structure

The practical effect of location dependence is that allocation probabilities vary systematically across the index. In the longitudinal VAR model,

$(s,t)$6

so two children with different baseline covariates have different prior probabilities of belonging to the same latent autoregressive regime (Beraha et al., 2022). The paper emphasizes that the prior is specified on the random partition of patients, which implies that posterior clusters can be driven by responses, covariates, or both; accordingly, “number of clusters” need not equal “number of distinct trajectory profiles” (Beraha et al., 2022).

In spatial and spatio-temporal stick-breaking, dependence is localized through kernel overlap. For the single-atom model,

$(s,t)$7

so nearby $(s,t)$8 and $(s,t)$9 have larger coincidence probability because the local kernels overlap more strongly (Grazian, 2023). This yields smooth index dependence without requiring Gaussian-process atoms in the baseline single-atom DDP construction. The same source notes, however, that single-atom DDPs have a positive lower bound on dependence; this motivates an extension with both $z$0 and $z$1 varying (Grazian, 2023).

A recurring conceptual distinction concerns whether dependence enters through weights or atoms. Linear-DDP-type competitors may let atoms depend on covariates while leaving weights common, whereas logit stick-breaking priors change the weights directly (Beraha et al., 2022). Spatial DDPs may instead keep the weights common and vary $z$2. The common-atoms, varying-weights form is canonical precisely because it changes local partition probabilities while preserving a shared global dictionary (Grazian, 2023, Horiguchi et al., 2022).

Tree topology further alters the prior dependence structure. The lopsided tree underlying ordinary sequential stick-breaking induces strong cross-covariate prior correlation and inherited stochastic ordering, whereas the balanced tree produces a symmetric, shallower path representation (Horiguchi et al., 2022). For a balanced tree with $z$3,

$z$4

so cross-location dependence can decay with depth, while local continuity is preserved if

$z$5

(Horiguchi et al., 2022). This directly links topology to the degree of prior smoothing across locations.

A related misconception concerns the meaning of “location.” In some papers it is spatial position; in others it is a predictor vector. The obesity application states this explicitly: the “location” index is the subject’s baseline covariate vector $z$6, not physical space (Beraha et al., 2022). The general notion is therefore index dependence in the mixing weights, not necessarily geostatistical locality.

4. Posterior computation and algorithmic strategies

The computational appeal of location-dependent stick-breaking priors lies in the fact that several important constructions reduce the dependent-weight problem to conditionally Gaussian updates.

The central device is Pólya–Gamma augmentation. In the LSBP model, the sequential logistic representation of the breaks yields binary regression likelihoods for the continuation-ratio probabilities, and introducing $z$7 makes the full conditional for $z$8 Gaussian (Rigon et al., 2017). The same idea is used in the finite VAR clustering model, where the full conditional for the weight parameters $z$9 is derived in closed form using auxiliary variables following Polson, Scott and Windle (2013) and Rigon and Durante (2021), again via Pólya–Gamma augmentation (Beraha et al., 2022). In finite multinomial logistic stick-breaking, the multinomial decomposes into $w_h(x)$0 binomial logistic terms,

$w_h(x)$1

so conditioned on $w_h(x)$2, the likelihood becomes Gaussian in $w_h(x)$3 (Linderman et al., 2015).

Truncation is another standard tactic. The obesity model uses finite truncation with $w_h(x)$4 in simulations and $w_h(x)$5 in the application (Beraha et al., 2022). LSBP likewise imposes $w_h(x)$6 in the truncated approximation (Rigon et al., 2017). The spatio-temporal stick-breaking process uses a conditional MCMC approximation with $w_h(x)$7 in simulations and applications (Grazian, 2023). These are blocked or conditional truncation schemes in the style of Ishwaran–James and related conditional samplers.

The latent-allocation step always couples local weights to local likelihood contributions. In the VAR model,

$w_h(x)$8

so covariate dependence affects clustering exactly through the weight factor (Beraha et al., 2022). In the LSBP Gaussian density-regression model,

$w_h(x)$9

and analogous local-allocation probabilities appear in spatial and tree-based models (Rigon et al., 2017, Grazian, 2023, Horiguchi et al., 2022).

Several constructions admit more than one inference regime. LSBP provides Gibbs sampling, expectation-maximization, and mean-field variational Bayes, all based on the same Pólya–Gamma augmentation (Rigon et al., 2017). The multinomial logistic stick-breaking model combines the augmentation with Gaussian-process regression, LDS smoothing, and block Gaussian updates (Linderman et al., 2015). TreeSB models preserve the same node-wise binary-regression decomposition under alternative tree topologies, so the move from lopsided to balanced trees does not destroy conditional-conjugate regression machinery (Horiguchi et al., 2022).

Other algorithmic strategies are more specialized. The spatio-temporal stick-breaking process uses latent Bernoulli variables

$$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$0

which separates the global beta variable $$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$1 from the local kernel activation probability $$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$2; knot variables $$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$3 and $$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$4 are then updated with Metropolis–Hastings (Grazian, 2023). This illustrates a broader pattern: local dependence in the weights often requires auxiliary-variable designs that decouple global break magnitudes from local gating.

5. Theoretical properties and adjacent models

A basic requirement is normalization. For LSBP, the weights satisfy

$$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$5

and the paper proves this under the condition that $$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$6 (Rigon et al., 2017). The same paper also gives an $$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$7 truncation bound: $$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$8 so the finite approximation error decays exponentially in $$F(\cdot)=\sum_{k=1}^{\infty}\pi_k\delta_{\theta_k}(\cdot), \qquad \pi_1=V_1,\quad \pi_k=V_k\prod_{j=1}^{k-1}(1-V_j),\ k\ge2,$$9 (Rigon et al., 2017).

For the spatio-temporal stick-breaking process, the weights are proper in the sense that

$\{\pi_k\}$0

provided $\{\pi_k\}$1 and $\{\pi_k\}$2 are positive (Grazian, 2023). Under these mild conditions, invoking Barrientos, Jara and Quintana (2012), the random measures are marginally DP-distributed for each $\{\pi_k\}$3, the process has full weak support, and the induced DP mixture model has smooth trajectories as $\{\pi_k\}$4 varies (Grazian, 2023).

Covariance structure is more subtle than the kernel form alone might suggest. A separable weight kernel,

$\{\pi_k\}$5

does not imply a separable covariance for the induced process because the stick-breaking product structure is nonlinear (Grazian, 2023). This is one of the clearest examples of how local dependence in weights produces nontrivial second-order behavior.

A broader theoretical backdrop comes from work that is not itself location-dependent. Exchangeable-length-variable stick-breaking processes generalize classical independent-stick priors and give properness and full-support criteria for dependent lengths, but they use one global weight sequence rather than $\{\pi_k\}$6-indexed weights (Gil-Leyva et al., 2020). Markov stick-breaking processes replace independent lengths by a Markov chain and show that dependence among $\{\pi_k\}$7 can be tuned while preserving prescribed marginals, again without introducing a location index (Gil-Leyva et al., 23 Jan 2026). These models are relevant background because they separate marginal stick laws from dependence mechanisms, a design principle that is directly transferable to location-dependent priors.

Two nearby constructions clarify what does not count, or counts only indirectly, as location-dependent stick-breaking. The multiscale Bernstein polynomial prior is a tree-structured stick-breaking density model with localized beta basis functions, but its weights are indexed by tree node $\{\pi_k\}$8, not by an external predictor or spatial coordinate (Canale et al., 2014). The $\{\pi_k\}$9-stick-breaking model for related samples couples mixture weights across discrete groups $$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$0 through shared and idiosyncratic sticks, but it is group-dependent rather than spatially or covariate-indexed, and its “location sensitivity” enters mainly through kernel perturbation of the atoms (Soriano et al., 2017). These contrasts help delimit the subject: location-dependent stick-breaking priors are specifically those in which the stick variables or the resulting weights vary with an external index.

6. Applications, advantages, and limitations

The applied literature shows that location-dependent weights are useful when cluster proportions or local mixture composition are expected to vary systematically with baseline factors, spatial position, or time. In the child-obesity application, the prior clusters children through subject-specific VAR coefficient matrices $$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$1, after adjusting for baseline fixed effects and a global growth trend; the authors report better predictive performance relative to a purely parametric model ($$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$2), a DP prior on $$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$3 without covariate dependence, and a model where covariates affect atoms linearly but not weights, and they state that the reported WAIC values favored their logit stick-breaking model (Beraha et al., 2022).

In density regression, the LSBP toxicology application models the conditional distribution of gestational age at delivery given maternal serum DDE concentration using $$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$4 observations, a Gaussian mixture kernel with

$$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$5

and a natural cubic spline basis in the stick-breaking logits with truncation $$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$6 (Rigon et al., 2017). The fitted conditional densities show increasing left-tail inflation with higher DDE, and Gibbs, EM, and VB give similar substantive results; EM is fastest for point estimation, VB is much faster than MCMC while retaining uncertainty quantification, and Gibbs gives exact posterior inference at greater computational cost (Rigon et al., 2017).

In spatial and spatio-temporal prediction, the empirical gains can be large. In one simulation, the benchmark separable Gaussian spatio-temporal model had ESPE $$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$7, the spatial stick-breaking competitor with temporally evolving atoms had $$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$8, and the proposed single-atom spatio-temporal stick-breaking had $$G_x=\sum_{h} w_h(x)\delta_{\theta_h}, \qquad w_h(x)=V_h(x)\prod_{\ell<h}\{1-V_\ell(x)\},$$9 (Grazian, 2023). On Australian rainfall data, using observations from 2007–2016 to predict 2017, the ESPEs were $x$00 for the Gaussian spatio-temporal model, $x$01 for spatial stick-breaking, and $x$02 for single-atom stSB; on California temperature data, the ESPEs were $x$03, $x$04, and $x$05, respectively (Grazian, 2023). The paper reports that spatial-only models oversmoothed, whereas stSB better captured local peaks, changing regional patterns, and recent local changes (Grazian, 2023).

Finite location-dependent stick-breaking on the simplex is likewise effective in structured multinomial settings. The multinomial GP and multinomial LDS models use latent Gaussian priors on stick logits $x$06 over year, latitude, longitude, or latent state trajectories, and the paper reports that the GP multinomial model is comparable in prediction to logistic-normal GP but considerably more efficient computationally, while the multinomial LDS is orders of magnitude faster than logistic-normal LDS with particle MCMC (Linderman et al., 2015).

Several advantages recur across these applications. Predictor-dependent clustering allows systematic shifts in prior cluster membership as covariates change (Beraha et al., 2022). Kernel-based localization allows local borrowing of strength while avoiding the oversmoothing of kriging-like Gaussian processes and the rigidity of purely spatial stick-breaking models that cannot adapt in time (Grazian, 2023). Binary-regression formulations preserve computational tractability through Pólya–Gamma augmentation (Rigon et al., 2017, Horiguchi et al., 2022).

The limitations are equally consistent. Truncation levels such as $x$07 or $x$08 must be chosen (Beraha et al., 2022, Grazian, 2023). Prior sensitivity can be substantial, especially for $x$09 and atom-variance hyperparameters; vague priors can lead to extreme imputations and poor mixing (Beraha et al., 2022). In weight-dependent clustering models, clusters may differ in covariate composition even when trajectory shapes are similar, because covariates directly enter the allocation probabilities (Beraha et al., 2022). Single-atom spatio-temporal DDPs retain a positive lower bound on dependence unless the atoms vary as well (Grazian, 2023). Logistic stick-breaking for multinomial probabilities is asymmetric, so category ordering matters (Linderman et al., 2015). Balanced finite trees mitigate several lopsided-stick pathologies, but infinite balanced-tree constructions are delicate and can degenerate without depth-dependent regularization (Horiguchi et al., 2022).

Taken together, these results establish location-dependent stick-breaking priors as a broad methodological class rather than a single model. Their unifying principle is simple: retain the stick-breaking normalization, but let the break probabilities vary with an external index. The resulting family encompasses predictor-dependent density regression, covariate-dependent longitudinal clustering, spatial and spatio-temporal DDPs, tree-structured gating models, and finite multinomial regressions. The primary modeling choice is therefore not whether to use stick-breaking, but where to place the dependence—in the weights, the atoms, or both—and how strongly to couple nearby locations through that dependence (Beraha et al., 2022, Grazian, 2023, Horiguchi et al., 2022).