---
title: Global Dependence Order Explained
url: https://www.emergentmind.com/topics/global-dependence-order
type: topic
---

# Global Dependence Order Explained

Global dependence order refers to formal mechanisms for ranking or ordering random variables or multivariate predictors according to the “strength” of their probabilistic dependence with a response variable. Approaches to global dependence ordering form crucial underpinnings in sensitivity analysis, model interpretability, and functional data analysis, enabling both principled comparison of input-output relationships and the calibration of dependence measures against explicit axioms. Two major frameworks embodying the concept are the conditional convex order (as a strict global order on functional dependence, satisfying a full suite of axioms) [2511.06498], and empirical global dependence orders based on sensitivity indices derived from dissimilarity-based dependence measures [1311.2483]. Both frameworks provide invariance properties, algorithmic ranking procedures, and operational interpretations in statistical modeling and machine learning.

## 1. The Conditional Convex Order as a Global Dependence Order

Let $(Y, X)$ and $(Y', X')$ be random vectors, with $Y, Y'$ real-valued and $X, X'$ having (potentially different) dimensions. The global dependence order, called the **conditional convex order** (“ccx-order”), is defined as follows:

- For $v \in (0,1)$, let $q_Y(v) = \inf \{y: F_Y(y) \geq v\}$ be the generalized inverse of the CDF $F_Y$, and denote $\eta_{Y|X}^v := \mathbb{P}(Y \geq q_Y(v)\;|\;X)$.
- Then $(Y, X) \preceq_{ccx} (Y', X')$ if for every $v \in (0,1)$,
  $$
  \eta_{Y|X}^v \leq_{cx} \eta_{Y'|X'}^v,
  $$
  meaning that for every convex function $\varphi : [0,1] \to \mathbb{R}$,
  $$
  \mathbb{E}[\varphi(\eta_{Y|X}^v)] \leq \mathbb{E}[\varphi(\eta_{Y'|X'}^v)].
  $$

This order compares the “spread” of conditional survival probabilities across predictors, capturing their joint ability to localize $Y$ given $X$. The approach is copula-invariant and compatible with dimension-reduction schemes.

## 2. Axiomatic Characterization: The Eight Axioms

A relation $\preceq$ on pairs $(Y, X)$ is a global dependence order if and only if it satisfies the following axioms:

| Code | Name                                  | Description                                                                                                                                    |
|------|---------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------|
| O1   | Law-invariance                        | If $(Y, X)$, $(Y', X')$ have identical joint laws, then they are equivalent in order: $(Y, X) \approx (Y', X')$.                              |
| O2   | Quasi-ordering                        | Reflexivity and transitivity hold: $\preceq$ is a quasi-order.                                                                                 |
| O3   | Independence as unique minimum         | $(Y, X) \preceq (Y', X')$ for all $(Y', X')$ if and only if $X \perp Y$.                                                                      |
| O4   | Perfect-dependence as unique maximum  | $(Y', X') \preceq (Y, X)$ for all $(Y', X')$ if and only if $Y = f(X)$ a.s. for some measurable $f$.                                          |
| O5   | Information monotonicity (data proc.) | For any predictor enlargement $Z$, $(Y, X) \preceq (Y, (X, Z))$.                                                                              |
| O6   | Conditional independence char.        | $(Y, X) \approx (Y, (X, Z))$ if and only if $Y \perp Z \mid X$.                                                                               |
| O7   | Transformation invariance             | $(Y, X) \approx (g(Y), h(X))$ for any strictly increasing $g$, bijection $h$.                                                                 |
| O8   | Copula (marginal) invariance          | Replacing $Y$ by $F_Y(Y)$ or $X$ by its marginal PIT leaves the order unchanged.                                                              |

These axioms uniquely determine the ccx-order among all possible orders of dependence strength. Of note is the information monotonicity (O5), which ensures that adding redundant predictors cannot decrease perceived dependence, and the *exclusive* conditional-independence characterization (O6), unattained in other orderings.

## 3. Extremal Elements: Independence and Perfect Dependence

The conditional convex order realizes independence and deterministic functional dependence as the universal minimum and maximum, respectively:

- **Minimal (Independence):** For all $v$, $\eta_{Y|X}^v$ is constant (unaffected by $X$), so $(Y, X)$ is minimal.
- **Maximal (Functional dependence):** For all $v$, $\eta_{Y|X}^v$ takes values only in $\{0,1\}$ a.s.; such two-point conditional variables are maximal in convex order.

This guarantees that $(Y, X)$ with $Y=f(X)$ is strictly greater than any partially dependent or independent system.

## 4. Fundamental Properties: Conditional Independence, Invariance, Data-processing

### Conditional Independence and Data-processing
For any $(Y, X, Z)$, the order guarantees $(Y, X) \approx (Y, (X, Z))$ if and only if $Y \perp Z\,|\,X$, making ccx-order uniquely sensitive to all conditional independences.

**Data-Processing Property:** For any additional predictor $Z$, $(Y, X) \preceq (Y, (X, Z))$. If $Z$ adds no information about $Y$ given $X$, equality holds.

### Invariance
- **Monotone Recoding:** Strictly increasing $g$ on $Y$’s support and bijections on $X$ do not affect order.
- **Copula Invariance:** Order is invariant under marginal probability integral transforms, highlighting the intrinsic copula-level comparison of the dependence structure.

## 5. Relationship to Classical Orders

### Schur Order and Concordance Order

- **Schur Order:** The ccx-order is equivalent to a Schur ordering of conditional CDFs:
  $$
  u \mapsto \eta_{Y|X}^v(q_X(u))
  $$
  is Schur-smaller than that for $(Y', X')$, i.e., has more concentrated partial integrals, in the sense of Hardy–Littlewood–Pólya.

- **Concordance Order ($\leq_c$):** For bivariate stochastically increasing (SI) models with continuous margins, $(Y, X) \preceq_{ccx} (Y', X')$ coincides with $(Y, X) \leq_c (Y', X')$, i.e., the pointwise order on joint CDFs and survival functions. More generally, concordance order’s extrema are comonotonic/anti-comonotonic copulas, whereas ccx-order’s maxima and minima are strict independence and functional dependence.

## 6. Instantiation in Key Model Classes

### Additive Error Models

For $Y=f(X)+\sigma \epsilon$, $\epsilon \perp X$, the ccx-order quantifies the effect of noise:
- If $0 \leq \sigma < \sigma'$, then $(Y, X) \succeq_{ccx} (Y', X)$.
- Any dependence measure increasing in ccx-order (e.g., Chatterjee’s $\xi$) is decreasing in $\sigma$.

### Multivariate Normal Distributions

For $(Y, X) \sim \mathcal{N}(\mu, \Sigma)$,
$$(Y, X) \preceq_{ccx} (Y', X') \iff \Sigma_{Y,X} \Sigma_X^{-1} \Sigma_{X,Y}/\sigma_Y^2 \leq \Sigma'_{Y,X} {\Sigma'_X}^{-1} \Sigma'_{X,Y}/\sigma_{Y'}^2$$
The classical fraction of explained variance $\operatorname{Var}[\mathbb{E}(Y|X)]/\operatorname{Var}(Y)$ is monotone in the ccx-order.

### Copula-Based Models

For $Y \perp X\, |\, g(X)$ and $Y' \perp X' \mid h(X')$, if the bivariate copulas $C_{Y, g(X)}$ and $C_{Y', h(X')}$ lie within a stochastically increasing family ordered by concordance, then $C_{Y, g(X)} \leq_c C_{Y', h(X')}$ implies $(Y, X) \preceq_{ccx} (Y', X')$.

## 7. Dependence Measures and the Global Dependence Order

### Chatterjee’s $\xi$

The rank correlation $\xi(Y, X)$ is defined as
$$
\xi(Y, X) = \frac{\int \operatorname{Var}(\mathbb{P}(Y \geq y\,|\,X))\, dP^Y(y)}{\int \operatorname{Var}(1_{Y \geq y})\, dP^Y(y)}
$$
Alternatively, $\xi(Y, X) \propto \int [\mathbb{E}(\eta_{Y|X}^v)^2]\, dv - \mathrm{const}$.
It is monotonic with respect to the ccx-order.

### Generalized $\phi$-functionals

For any convex $\phi$, define
$$
\xi_{\phi}(Y, X) = \alpha_{\phi}^{-1} \int \int \phi(F_{Y|X=x}(y) - F_Y(y))\, dP^X(x) dP^Y(y)
$$
Two-point “sensitivity” versions $\Lambda_{\phi}$ similarly compare $\phi$ of pairwise differences of conditional CDFs.
All such measures are ccx-increasing, vanishing only at independence, and attaining their maximum at perfect dependence.

### Rearranged Concordance-Based Measures

For any copula-based concordance measure $\mu$ on SI vectors (such as Spearman’s $\rho$, Gini’s $\gamma$), define
$$
\mu(Y, X) := \mu(F_Y \circ q_{Y|X}^{\uparrow}(V), U)
$$
where $q_{Y|X}^{\uparrow}$ is the increasing rearrangement by $V \sim \mathrm{Uniform}(0,1)$.
All such $\mu$ measures are ccx-increasing, characterize independence and perfect dependence, and possess information-monotonicity.

## 8. Empirical Global Dependence Orders from Sensitivity Indices

Alternative approaches establish a global dependence order on input variables via sensitivity index-based ranking [1311.2483]:

Let $Y = \eta(X^1, \ldots, X^p)$, with independent $X^k$ and known law. For each input $X^k$, define:
$$
S_k := \mathbb{E}_{X^k} [ d(Y,\,Y|X^k) ]
$$
where $d(\cdot, \cdot)$ is a suitable dissimilarity or divergence. Key realizations include:

- **f-divergence indices:** Use convex $f$ to measure divergences between $p_Y$ and $p_{Y|X^k=x}$.
  - **Borgonovo’s index:** $f(t) = |t-1|$
  - **Mutual information:** $f(t) = t\ln t$
  - **Pearson $\chi^2$:** $f(t) = (t-1)^2/t$
- **Distance correlation ($\mathrm{dCor}$):** Characteristic function–based quadratic integral, normalized to $[0,1]$.
- **Hilbert–Schmidt Independence Criterion (HSIC):** Kernel-trace based covariance on RKHS.

A total ordering results from ranking $S_k$ (or its estimator) across variables. Procedures include sorting, mRMR for redundancy-penalized selection, or feature selection via HSIC-Lasso.

## 9. Computation and Practical Illustration

For f-divergence indices, density ratio estimation (via unconstrained least-squares importance fitting or ML) is used. dCor and HSIC are both $O(n^2)$ in sample size and provide pseudocode for efficient computation. All indices are:

- Transformation invariant (to varying degrees)
- Non-additive (do not sum to one as variance-based Sobol indices)
- Consistent as $n \to \infty$
- Able to handle multivariate $X^k$ and $Y$

Empirical global dependence orders outperform Sobol indices in nonlinear, interaction-heavy, or high-dimensional settings, as in the Ishigami and Morris models or industrial cases (e.g., Punq reservoir, Marthe maps). This suggests their robustness in screening influential inputs under minimal structural assumptions.

## 10. Summary and Outlook

A global dependence order provides a law-invariant, data-processing-respecting framework to order (possibly multivariate) variables by their dependence strength with an output. The conditional convex order is characterized by a unique set of axioms and connects to Schur and concordance orderings, while also providing operational semantics for a variety of dependence measures. Empirical approaches based on generalized sensitivity indices yield practical ordering and screening tools in complex, high-dimensional, and nonlinear models, underscoring the central role of robust dependence quantification in both theory and application.

Source: https://www.emergentmind.com/topics/global-dependence-order