---
title: Global Moran’s I in Spatial Analysis
url: https://www.emergentmind.com/topics/global-moran-s-i
type: topic
---

# Global Moran’s I in Spatial Analysis

Global Moran’s I is a quadratic-form statistic that serves as the classical global index of spatial autocorrelation for a univariate attribute distributed over a spatial domain. Its rigorous mathematical construction, spectral properties, relationships to alternative spatial statistics, and role in spatial econometric modeling make it a foundational tool in modern spatial analysis.

## 1. Definition and Mathematical Structure

Let $x = (x_1, \ldots, x_N)^\top$ be a real-valued variable observed on $N$ spatial units, with sample mean $\bar x$ and (possibly sample or population) variance $\sigma_x^2 = \sum_{i=1}^N (x_i - \bar x)^2/N$. Fix a spatial-weight matrix $W = [w_{ij}]$ with zero diagonal ($w_{ii}=0$) and $\sum_{i,j} w_{ij} = 1$ (global normalization). The z-score standardized vector is $z = (x - \bar x \mathbf{1})/\sigma_x$, satisfying $z^\top \mathbf{1} = 0$, $z^\top z = N$.

The global Moran’s I is defined as
$$
I = \frac{\sum_{i=1}^N \sum_{j=1}^N w_{ij} (x_i - \bar x)(x_j - \bar x)}{\sum_{i=1}^N (x_i - \bar x)^2}
= \frac{z^\top W z}{z^\top z} ,
$$
which, under global normalization and $z^\top z=1$, reduces to $I = z^\top W z$ [2508.19908].

Moran’s I quantifies the degree of spatial correlation: $I>0$ signals clustering of similar values, $I<0$ anti-clustering (checkerboard-like alternation), and $I\approx 0$ spatial randomness.

## 2. Spectral Properties, Bounds, and Interpretation

Moran’s I can be interpreted as a Rayleigh quotient, with attainable values determined by the spectral properties of $W$. For symmetric $W$, extremal values satisfy
$$
\lambda_{\min}(W) \leq I \leq \lambda_{\max}(W) ,
$$
where $\lambda_{\min}$ and $\lambda_{\max}$ are the smallest and largest eigenvalues of $W$ (projected onto the zero-mean subspace) [1501.06260, 2209.08562]. If $W$ is globally normalized, the “absolute” theoretical bounds are $-1 < I < 1$, although in practice the achievable range is $n\lambda_{\min} \leq I \leq n\lambda_{\max}$, intersected with $[-1,1]$ [2209.08562]. Row-standardization of $W$ can enforce $I \in [-1,1]$, but general weighting schemes (including irregular degree graphs or distance-based weights) may yield I-values outside the unit interval [2112.10708].

A monotone transformation of $I$ to $I^* = ((n-1)I + 1)/C$ rescales it into $[-1,1]$, where $C$ is computed from the spectral bounds, yielding a direct interpretational comparability to Pearson’s correlation [1501.06260].

## 3. Structural Decomposition and Connection with Other Indices

Chen introduced an exact decomposition of global Moran’s I in terms of the Getis–Ord indices and the size correlation function [2508.19908]:
$$
I = \frac{N^2 G - 2 N G_L + 1}{N C_f - 1} ,
$$
where
- $p_i = x_i / \sum_{k=1}^N x_k$ is the normalized size vector,
- $C_f = \sum_{i=1}^N p_i^2$ is the size-correlation function,
- $G = p^\top W p$ is the global Getis–Ord index,
- $G_L = \sum_{i=1}^N G_i$, $G_i = (W p)_i$ is the sum of local Getis–Ord indices.

This structural decomposition reveals that global spatial autocorrelation depends on four distinct components: the global Getis–Ord index, the sum over local Getis–Ord indices, the number of spatial elements $N$, and $C_f$—the correlation structure of the normalized attribute. The local Getis–Ord indices are equivalent to gravity-model–based potential indices, establishing the association of Moran’s I with spatial interaction: weak spatial interaction implies weak global autocorrelation [2508.19908]. The relationship is strictly nonlinear, and the interplay between $G$ and $G_L$ (and, implicitly, the gravity model) governs the magnitude and sign of $I$.

## 4. Algebraic, Regression, and Eigenvalue Perspectives

Moran’s I admits multiple algebraic forms:
- Quadratic form: $I = z^\top W z$;
- Outer-product eigen formulation: $(z z^\top) W z = I z$ (so $I$ is a nonzero eigenvalue of $z z^\top W$);
- Inner-product formulation: $n W z = a \mathbf{1} + I z + \epsilon$, providing regression-model–based interpretations [2209.08560, 2203.13188].

Regression techniques can compute $I$ as the slope of the least-squares fit of $n W z$ on $z$, either with or without intercept, and thus directly connect autocorrelation measurement with spatial autoregressive models. The inner-product equation $n W z = I z$ is formally the inverse of the simplest noise-free spatial autoregressive (SAR) model $z = \rho W z$, with the explicit relationship $I = n/\rho$ in that limiting case [2203.13188]. This reconciliation aligns spatial autocorrelation estimation and the SAR coefficient within a unified algebraic framework and clarifies stability domains for SAR estimation via eigen-analysis [2203.13188, 2209.08560].

## 5. Weight Matrix Choices and Their Implications

The choice and normalization of the weight matrix $W$ are fundamental to the behavior of global Moran’s I [1606.03658, 2112.10708]. Standard options include:
- Binary adjacency (e.g., rook or queen contiguity in planar maps),
- Row-standardized versions ($P$) so that rows sum to one,
- Inverse distance decay ($w_{ij} \propto 1/d_{ij}$ or $w_{ij} \propto \exp(-d_{ij}/h)$),
- Graph Laplacian ($L = D-A$), which inverts the interpretation of I,
- Doubly-stochastic matrices ($M$), e.g. Metropolis–Hastings random-walk kernels for robust boundedness $I \in [-1,1]$ and interpretable Markov variance-retention properties [2112.10708].

The spectrum of $W$ and its degree of regularity (homogeneity of neighbor counts) directly control the attainable range of $I$ and the stability of its interpretation [2112.10708, 1501.06260]. Irregular graphs can yield arbitrarily extreme $I$ values for suitably concentrated attribute vectors.

## 6. Statistical Inference and Information-Theoretic Interpretation

Under the null hypothesis of spatial randomness (random permutations of spatial labels), the expected value is $E[I] = -1/(n-1)$ [2401.05728, 1710.04549]. Analytical variance expressions exist, depending on $W$ and higher moments of $x$. Both parametric (normal approximation) and permutation-based inference for $I$ are standard: the latter matches the exact reference distribution under exchangeability [2012.08647, 2110.10622]. Permutation-free closed-form p-value bounds using concentration inequalities have been developed for computational efficiency at scale [2012.08647, 2110.10622].

Recent developments connect Moran’s I to self-information (surprisal) and entropy. The probability of observing a given value $I = \alpha$ under random spatial arrangement defines $J(\alpha) = -\log P(I=\alpha)$—yielding an information-theoretic measure of spatial surprisingness. High autocorrelation (high $I$) corresponds to high self-information and low entropy, i.e., the data are compressible and “surprising” given the underlying spatial arrangement [2405.18459]. This provides a bridge between spatial statistics and information theory, and motivates loss functions in spatial learning or ecological inference.

## 7. Extensions, Applications, and Analytical Generalizations

Analytical generalizations include scale-dependent spatial autocorrelation functions (SACF), constructed by parametrizing $W$ with a spatial lag $h$ (e.g., “relative step” function) [2001.06750]. This yields cumulative and density SACFs, analogous to temporal autocorrelation functions, and admits partial SACF versions via spatial Yule–Walker equations. The resulting framework allows multi-scale and fractal analysis of spatial autocorrelation, with implications for identifying correlation dimensions in geographical systems [2001.06750].

Empirical studies demonstrate the versatility of global Moran’s I: from quantifying the tempo-spatial clustering of infectious disease [2011.01739], measuring spatial balance in sample surveys [1710.04549], to providing a null-preserving resampling algorithm that fixes observed autocorrelation for hypothesis testing [2401.05728]. The index is foundational for exploratory spatial data analysis, spatial panel model diagnostics, and for distinguishing clustering, randomness, and dispersion in heterogeneous spatial systems [1606.03658, 2110.10622].

---

**References**  
[2508.19908]: "Structural Decomposition of Moran's Index by Getis-Ord's Indices"  
[2112.10708]: "Measuring Segregation via Analysis on Graphs"  
[2405.18459]: "Probing the Information Theoretical Roots of Spatial Dependence Measures"  
[2401.05728]: "A General Method for Resampling Autocorrelated Spatial Data"  
[1606.03658]: "New Approaches for Calculating Moran's Index of Spatial Autocorrelation"  
[2001.06750]: "An Analytical Process of Spatial Autocorrelation Functions Based on Moran's Index"  
[1501.06260]: "An alternative to Moran's I for spatial autocorrelation"  
[2011.01739]: "Spatial autocorrelation and the dynamics of the mean center of COVID-19 infections in Lebanon"  
[2209.08560]: "Spatial autocorrelation equation based on Moran's index"  
[2209.08562]: "Deriving two sets of bounds of Moran's index by conditional extremum method"  
[2012.08647]: "Computation-free Nonparametric testing for Local and Global Spatial Autocorrelation with application to the Canadian Electorate"  
[2110.10622]: "Local Statistics for Spatial Panel Models with Application to the US Electorate"  
[2203.13188]: "Derivation of an Inverse Spatial Autoregressive Model for Estimating Moran's Index"  
[1710.04549]: "Measuring the spatial balance of a sample: A new measure based on the Moran's I index"

Source: https://www.emergentmind.com/topics/global-moran-s-i