---
title: Multiscale Similarity GWR (M-SGWR) Model
url: https://www.emergentmind.com/topics/multiscale-similarity-gwr-m-sgwr
type: topic
---

# Multiscale Similarity GWR (M-SGWR) Model

Multiscale Similarity Geographically Weighted Regression (M-SGWR) is a spatial regression framework that extends the classical geographically weighted regression (GWR) and its multiscale generalizations. M-SGWR explicitly quantifies local regression relationships using both spatial proximity and attribute (covariate) similarity, allowing different predictors to exhibit varying degrees of dependence on geographic versus contextual, non-contiguous similarity. This dual weighting addresses limitations of conventional GWR/MGWR models in settings where local similarity is not purely spatially contiguous, as in network diffusion or socio-demographic contagion. The model was formally introduced by Lessani et al. [2601.19888].

## 1. Mathematical Foundations and Model Specification

Given a response vector $y\in\mathbb{R}^n$ and predictors $X=[\mathbf 1, \mathbf x_1, \ldots, \mathbf x_{m-1}]\in\mathbb{R}^{n\times m}$ at spatial locations $(u_i, v_i)$, M-SGWR fits local weighted least-squares (WLS) models at each site $i$. Each predictor $j$ is associated with two key parameters:

- Bandwidth $b_j$ specifying the spatial scale for geographic proximity.
- Mixing parameter $\alpha_j \in [0,1]$ quantifying the balance between geographic and attribute-based weighting.

For each location $i$ and covariate $j$, two diagonal weight matrices are constructed:
- **Geographic kernel**: $W^{\rm geo}_{i,j}$ uses a fixed bandwidth $b_j$:
  $$
  w_{ij}^{\rm geo}(\ell) = \begin{cases}
    [1 - (d_{i\ell}/b_j)^2]^2 & \text{if } d_{i\ell} < b_j \\
    0 & \text{otherwise} \end{cases}
  $$
  with $d_{i\ell} = \|(u_i, v_i) - (u_\ell, v_\ell)\|$.
- **Attribute similarity kernel**: $W^{\rm attr}_{i,j}$, defined over geographic neighbors $N_i = \{\ell : w_{ij}^{\rm geo}(\ell)>0\}$:
  $$
  w_{ij}^{\rm attr}(\ell) =
    \begin{cases}
      \exp\left(-\left(\frac{x_{\ell j}-x_{ij}}{\mathrm{SD}_{i,j}}\right)^2\right) & \ell \in N_i \\
      0 & \text{otherwise}
    \end{cases}
  $$
  with $\mathrm{SD}_{i,j} = \mathrm{std} \{ x_{\ell j}: \ell \in N_i\}$.

The combined weight matrix for covariate $j$:
$$
W_{i,j}(\alpha_j) = \alpha_j W^{\rm geo}_{i,j} + (1-\alpha_j) W^{\rm attr}_{i,j}
$$

Local parameter estimates $\hat{\boldsymbol\beta}_i$ are computed as:
$$
\hat{\boldsymbol\beta}_i = (X^T W_i X)^{-1} X^T W_i \mathbf y
$$
with $W_i = \mathrm{diag}\{w_{i1}, \ldots, w_{in}\}$ and $w_{i,\ell} = \sum_{j=0}^{m-1} w_{ij}(\ell; \alpha_j, b_j)$.

## 2. Parameter Selection and Optimization

Efficient estimation in M-SGWR requires the determination of both $b_j$ and $\alpha_j$ for each predictor.

### 2.1 Model Selection Criteria

The two primary objective functions used are:
- **Corrected Akaike Information Criterion (AICc):**
  $$
  \mathrm{AICc} = n\ln(\hat\sigma^2) + n\ln(2\pi) + n\frac{n+\mathrm{tr}(S)}{n-\mathrm{tr}(S)-2}
  $$
  where $\hat{\sigma}^2$ is the WLS residual variance and $S$ is the hat-matrix.
- **Cross Validation (CV):**
  $$
  \mathrm{CV} = \frac{1}{n}\sum_{i=1}^n \left(\frac{e_i}{1 - S_{ii}}\right)^2
  $$
  with $e_i = y_i - \hat y_i$.

### 2.2 Iterative Backfitting

Simultaneous optimization across all predictors is performed via an iterative backfitting routine:
- Initialize $b_j$ and $\alpha_j$ for all $j$.
- For each predictor, hold others fixed, optimize $(b_j, \alpha_j)$ by minimizing AICc or CV.
- Update WLS estimates using new weights.
- Iterate until convergence (e.g., residual sum-of-squares or coefficient change).

This approach efficiently explores the combinatorial search space by reducing the high-dimensional problem into a series of one-dimensional optimizations, such as via golden-section or greedy search strategies.

## 3. Theoretical Properties and Model Generalization

M-SGWR encompasses several established local regression models:
- Setting all $\alpha_j = 1$ and equal $b_j$ recovers standard GWR.
- Allowing each $b_j$ but setting all $\alpha_j = 1$ yields MGWR.
- Fixed $b_j$, shared global $\alpha_j$ is equivalent to SGWR.

The parameter $\alpha_j$ provides an explicit control: as $\alpha_j \rightarrow 1$, predictor $j$’s kernel becomes purely geographic; as $\alpha_j \rightarrow 0$, attribute similarity dominates, supporting remote (non-contiguous) similarity. This generalization enables modeling of processes where spatial interaction is defined by contextual or attribute matching, not just physical proximity.

A plausible implication is that M-SGWR is structurally capable of capturing both spatially contiguous and “network-like” contextual dependencies, accommodating complex patterns in spatial social or epidemiological data.

## 4. Simulation Studies

M-SGWR has been empirically compared to MGWR under simulated scenarios:

- **Mixed “Geographic + Contextual” Effects (n=1200):**
  $y_i = \beta_{0}(u_i, v_i) + \beta_{1}(u_i, v_i) x_{i1} + \beta_{2}(u_i, v_i) x_{i2} + \varepsilon_i$, with $\beta_j$ comprising both contiguous and non-contiguous (contextual) spatial fields. 
  - M-SGWR yields lower RMSE and higher Pearson correlation for local $\beta$ estimation than MGWR.
  - Estimated $(\alpha_0, \alpha_1, \alpha_2) = (1.0, 0.489, 0.066)$ demonstrates that $\beta_2$ is primarily driven by attribute similarity.

- **Pure Geographic Effects:**
  All $\beta_j(u_i, v_i)$ are smooth spatial fields. M-SGWR with $\alpha_j \equiv 1$ reproduces MGWR coefficients exactly.

**Goodness-of-fit metrics used include:** RMSE for local $\beta$ estimates, adjusted $R^2$, AICc, residual sum-of-squares (RSS), and residual Moran’s $I$. 

| Model         | RMSE(β₀,β₁,β₂)                 | Pearson(β₀,β₁,β₂)            |
|---------------|-------------------------------|------------------------------|
| MGWR          | (0.260, 0.385, 0.544)         | (0.971, 0.766, 0.704)        |
| M-SGWR        | (0.224, 0.361, 0.511)         | (0.972, 0.798, 0.744)        |

## 5. Empirical Application: COVID-19 Analysis

An application to COVID-19 confirmed case rates across 616 counties in seven southern U.S. states involved covariates such as %Black, %Hispanic, %Bachelor’s, median income, %65+, %18–29, population density, %foreign born, and %uninsured.

### 5.1 Comparative Model Performance

| Model   | Adj. $R^2$ | AICc   | RMSE   |
|---------|------------|--------|--------|
| MGWR    | 0.701      | 1093.5 | 0.513  |
| M-SGWR  | 0.782      | 943.2  | 0.433  |

M-SGWR achieves the highest adjusted $R^2$, lowest AICc, and lowest RMSE. Residual spatial autocorrelation (Moran’s $I$) is also minimized under M-SGWR.

### 5.2 Interpretation of Estimated Parameters

Optimized $\alpha_j$ values reveal the role of attribute similarity in different predictors:
- $\alpha_{\rm Hispanic}=0.143$
- $\alpha_{\rm pop\,density}=0.024$
- $\alpha_{\rm income}=0.205$
- $\alpha_{\rm 65+}=0.052$

Low $\alpha_j$ (e.g., population density) indicates significant influence from attribute similarity, enabling the model to identify shared dynamics across non-local counties. High $\alpha_j$ values indicate primarily spatially contiguous effects.

## 6. Implications and Summary

M-SGWR generalizes previous geographically weighted regression models by allowing each covariate to select both its spatial scale $b_j$ and its own mix $\alpha_j$ of geographic versus attribute-based weighting. The methodology retains the interpretability and computational structure of local WLS regression while expanding the scope of spatial interaction to include both traditional spatial and non-spatial (attribute-based, network) similarity. In both synthetic and real spatial datasets, including epidemiological contexts, M-SGWR demonstrates superior fit and ability to model complex spatial-connectivity structures compared to GWR, SGWR, and MGWR [2601.19888].

Source: https://www.emergentmind.com/topics/multiscale-similarity-gwr-m-sgwr