---
title: Azadkia-Chatterjee Coefficient
url: https://www.emergentmind.com/topics/azadkia-chatterjee-coefficient
type: topic
---

# Azadkia-Chatterjee Coefficient

The Azadkia–Chatterjee coefficient is a nonparametric, rank-based measure of directed dependence between a vector-valued predictor and a univariate or multivariate response, defined at the population level via the variance of conditional probabilities and estimated using nearest-neighbor graphs. It features an interpretable scale—zero under independence and one under functional dependence—and a graph-based empirical estimator that admits parametric rates, strong consistency, bandwidth-free implementation, and central limit theorems in both marginal and conditional versions. Multivariate extensions, scale-invariant variants, and connections to broader classes of geometric graph and kernel-based dependence measures position the coefficient as a central object for independence testing, graphical models, and model-free variable selection.

## 1. Definition and Fundamental Properties

Let $(X,Y)$ be jointly distributed random elements with $X\in\mathbb{R}^d$ and $Y$ either univariate or a vector in $\mathbb{R}^{q}$. The **Azadkia–Chatterjee (AC) coefficient** for $Y$ on $X$ is defined by
\[
\xi(Y,X) \;=\; \frac{ \int_{\mathbb{R}} \operatorname{Var}\big( P(Y\ge y\mid X) \big) \, dP^Y(y) }{ \int_{\mathbb{R}} \operatorname{Var}\left( \mathbf{1}\{ Y \ge y \} \right) dP^Y(y) } \in [0,1].
\]
An equivalent form based on the cumulative distribution of $Y$ yields, for continuous $F_Y$:
\[
\xi(Y, X) = 6\int_\mathbb{R} \operatorname{Var}\left( P(Y\ge y \mid X) \right) dP^Y(y) - 2.
\]
**Characterizing properties**:
- $\xi(Y,X)=0$ if and only if $X$ and $Y$ are independent.
- $\xi(Y,X)=1$ if and only if $Y$ is almost surely a measurable function of $X$.
  
The definition is directional and scale-invariant: strictly increasing transformations of $Y$ or bijections of $X$ preserve $\xi$ [2503.11390][2212.01621]. For conditional dependence, define $(X,Y,Z)$ jointly and set
\[
\xi = \frac{ \int \operatorname{Var}\!\left( \mathbb{E}[\, \mathbf{1}\{ Y \ge y \} | X,Z\,] \,|\, X \right) dF_Y(y)} { \int \operatorname{Var}\!\left( \mathbf{1}\{ Y \ge y \} | X \right) dF_Y(y)}.
\]
$\xi = 0$ if and only if $Y \perp Z \mid X$, and $\xi = 1$ if and only if $Y$ is a function of $(Z,X)$ given $X$ [2108.06827][2012.14804].

## 2. Graph-Based and Rank-Based Estimator Construction

For i.i.d. data $\{ (X_i, Y_i) \}_{i=1}^n$, construct the following **graph-based estimator**:
- Compute the univariate ranks $R_i = \#\{j : Y_j \le Y_i\}$.
- Let $N(i) = \arg\min_{j\ne i} \| X_j - X_i \|$.
- The empirical AC coefficient is
  \[
  \xi_n = \frac{6}{n^2-1} \sum_{i=1}^n \min(R_i, R_{N(i)}) - \frac{2n+1}{n-1}.
  \]
This estimator generalizes Chatterjee's original proposal to multivariate covariates $X$ by utilizing nearest-neighbor graphs in $\mathbb{R}^d$ [2204.08031].

**Multivariate response**: For $Y\in \mathbb{R}^q$, a "chain rule" or copula-based construction is used [2212.01621][2512.07443]:
\[
T(Y|X) = \frac{\sum_{i=1}^q \left( \xi(Y_i | (X,Y_{<i})) - \xi(Y_i| Y_{<i}) \right)}{ \sum_{i=1}^q (1 - \xi(Y_i | Y_{<i})) },
\]
with $T(Y|X) \in [0,1]$, reduces to $\xi$ for $q=1$, and can be strongly consistently estimated using graph-based estimators for each univariate constituent.

**Scale invariance**: The standard estimator is not invariant to affine changes in $X$; a fully scale-invariant version uses coordinatewise rank transforms in $X$ before constructing the NNG [2412.02668].

## 3. Distributional Properties and Limit Theory

### Asymptotic Normality and Variance Bounds

The central limit theorem holds under broad conditions. For i.i.d. draws from a continuous law:
\[
\frac{\xi_n - \mathbb{E}[\xi_n]}{\sqrt{ \mathrm{Var}[\xi_n] }} \xrightarrow{d} N(0,1),
\]
whenever $Y$ is not a measurable function of $X$ [2204.08031]. The asymptotic variance $V_n = n\cdot \mathrm{Var}[\xi_n]$ satisfies:
\[
0 < \liminf_n V_n \leq \limsup_n V_n \leq 36,
\]
and, under absolute continuity of $F_X$, a sharper bound involving explicit dimension-dependent constants.

When $X \perp Y$, $\sqrt{n} \xi_n \xrightarrow{d} N(0, \tfrac{2}{5} + \tfrac{2}{5}\kappa_q + \tfrac{4}{5}\kappa_o )$ with $\kappa_q$, $\kappa_o$ linked to the geometry of the NNG in $\mathbb{R}^d$ [2204.08031][2209.11156]. Under manifold support, the limiting variance depends solely on the intrinsic dimension.

A **consistent explicit estimator** of the variance is available, allowing for valid inference [2204.08031].

### Symmetric and Conditional Extensions

A symmetrized version, taking $\max\{\xi_n(X,Y), \xi_n(Y,X)\}$, allows construction of two-sided tests—its limit law under independence is skew-normal with explicit variance [2205.01769].

The **conditional AC coefficient** admits an empirical estimator with parallel asymptotics; under independence, $\sqrt{n} \xi_n$ is asymptotically normal with variance determined by the dimensions of the variables and graph-count statistics [2108.06827].

### Continuity Considerations

Unlike classical measures (Spearman's rho, Kendall's tau), $\xi$ is not weakly continuous under distributional convergence. Instead, it is continuous with respect to convergence of Markov products (pairs of conditionally i.i.d. copies) under additional marginal quantile convergence or specific copula convergence. Practical families and models (ellipticals, Archimedeans, noises) satisfy required continuity, so stable large-sample inference is possible within these classes [2503.11390].

## 4. Algorithmic and Computational Aspects

- **Nearest-neighbor graph construction** can be done in $O(n\log n)$ (brute force for small $d$; kd-trees or approximate methods for larger $d$).
- **Rank computations** for $Y$ (and optionally for $X$ in the scale-invariant version) cost $O(n\log n)$ per coordinate.
- **Multivariate response**: Efficient merge-sort or divide-and-conquer algorithms exist for blockwise rank counts, with time complexity $O(n (\log n)^{q})$ [2512.07443].
- For each observation $i$, nearest-neighbor search and rank calculations admit nearly linear scaling, enabling use in large datasets.

## 5. Connections to Broader Dependence Measures

The AC coefficient is a specific instance within the family of graph–RKHS–OT dependency measures [2010.01768][2411.13080]:
- **Population level**: For sufficiently rich kernels (e.g., the min kernel on $[0,1]$, or the indicator-integral kernel), the corresponding normalized conditional MMD directly recovers $\xi$.
- **Sample level**: The estimator is a geometric graph functional over empirical OT ranks.
- **Distribution-free**: Under the null of independence, the law of the AC coefficient (when computed using empirical OT ranks and graph structure) is exactly permutation invariant, enabling finite-sample calibration for independence tests.

Multivariate extensions (both in predictors and responses) and conditional variants fit naturally into this graph–kernel framework, relating directly to kernel partial correlation [2012.14804], distance multivariance, and more general measures indexed by RKHS [2411.13080].

## 6. Practical Application Domains

### Independence and Conditional Independence Testing

The AC coefficient and its conditional extension are used for:
- Testing independence in arbitrary dimensions (direct, distribution-free under the null, with consistent critical values).
- Conditional independence testing, e.g., through graph-based statistics evaluated with (conditional) randomization tests [2108.06827]. However, these are known to exhibit low local power against contiguous local alternatives unless the nearest-neighbor graph is appropriately generalized or replaced with $k$-NN approaches.

### Graphical Model Structure Learning

Pairwise conditional AC coefficients are used as entries in adjacency matrices for learning undirected graphs representing conditional independence relationships in high dimensions, outperforming standard penalized Gaussian graphical model approaches in various regimes [2309.09082].

### Model-Free Feature Selection and Network Analysis

The multivariate T extension and its estimator enable:
- Directional, scale-invariant variable selection in high-dimensional regression settings [2212.01621][2503.11390].
- Ranking and forward feature selection for multivariate outcomes, with no tuning parameters and explicit stopping rules.
- Directed network inference in financial, biological, and climatological data [2212.01621].

## 7. Theoretical Limitations and Open Problems

- Under local parametric or minimax-detection boundary alternatives, the standard 1-NN estimator is asymptotically powerless unless graph construction is strengthened (increasing $k$ with $n$) [2108.06827].
- Weak continuity of $\xi$ fails under convergence in law, but holds under stricter Markov-product and copula-derivative types of convergence, implying care is needed in statistical inference [2503.11390].
- In practical high-dimensional settings, the curse of dimensionality in nearest-neighbor search may be partially circumvented due to intrinsic dimension adaptivity, but further analysis on computational–statistical tradeoffs remains ongoing [2209.11156].

---

**References:**
- [2204.08031] Limit theorems of Chatterjee's rank correlation
- [2205.01769] On the asymptotic distribution of the symmetrized Chatterjee's correlation coefficient
- [2108.06827] On Azadkia-Chatterjee's conditional dependence coefficient
- [2412.02668] On a rank-based Azadkia-Chatterjee correlation coefficient
- [2512.07443] A multivariate extension of Azadkia-Chatterjee's rank coefficient
- [2212.01621] A direct extension of Azadkia & Chatterjee's rank correlation to multi-response vectors
- [2209.11156] Azadkia-Chatterjee's correlation coefficient adapts to manifold data
- [2010.01768] Measuring Association on Topological Spaces Using Kernels and Geometric Graphs
- [2012.14804] Kernel Partial Correlation Coefficient -- a Measure of Conditional Dependence
- [2309.09082] A construction of a graphical model
- [2503.11390] On continuity of Chatterjee's rank correlation and related dependence measures
- [2411.13080] Distribution-free Measures of Association based on Optimal Transport

Source: https://www.emergentmind.com/topics/azadkia-chatterjee-coefficient