learn_spatial_autocorrelation

learn_spatial_autocorrelation(
    side=12,
    rho=0.6,
    n_sims=10,
    permutations=199,
    seed=0,
)

See what spatial autocorrelation looks like — and how Moran’s I tracks it.

Simulates fields y = (I - rho W)^-1 eps on a side x side lattice with row-standardized queen weights, sweeping the planted dependence ρ over a grid that includes the focal rho. The figure pairs the focal simulated field (left) with the Moran’s I recovered at each planted ρ (right): at ρ = 0 the statistic sits at its null expectation E[I] = -1/(n-1); as ρ rises, neighbors look alike and I climbs.

Parameters

Name	Type	Description	Default
side	int	Lattice side length (n = side²).	12
rho	float	The focal planted spatial dependence, |ρ| < 1. The left panel draws a field at this value and the sweep curve highlights it.	0.6
n_sims	int	Simulated fields per ρ (the faint markers behind the mean curve).	10
permutations	int	Conditional permutations behind each Moran’s I pseudo p-value.	199
seed	int	Random seed.	0

Returns

Name	Type	Description
	SandboxResult	df (one row per ρ and simulation), fig, summary, topic and the focal simulated field in data.

Examples

The knob variation is the lesson — compare no dependence with strong dependence:

import geometrics as gm

gm.learn_spatial_autocorrelation(rho=0.0).fig
gm.learn_spatial_autocorrelation(rho=0.8).fig