activeinferenceinstitute/geo_infer/geo-infer-spm

GEO-INFER-SPM

Instructions

Core Capabilities

• GLM fitting: General linear models with spatial design matrices
• Random field theory: Multiple comparison correction for spatial data
• Cluster inference: Cluster-level and peak-level statistics
• Contrast testing: T-contrasts and F-contrasts on spatial maps
• Visualization: Interactive time series explorer (mean±SD + residuals)

Key Imports

from geo_infer_spm.core.glm import GeneralLinearModel
from geo_infer_spm.core.rft import RandomFieldTheory
from geo_infer_spm.models.data_models import SPMData, SPMResult
from geo_infer_spm.visualization.interactive import create_time_series_explorer

Random Field Theory Inference

import numpy as np
from geo_infer_spm.core.rft import RandomFieldTheory
rft = RandomFieldTheory(
    field_shape=(64, 64),
    smoothness=np.array([4.5, 4.5]),
)
rft.compute_resel_counts()
peak_height = rft.peak_threshold(0.05, stat_type="Z", two_sided=True)
cluster_p = rft.cluster_extent_p_value(
    extent=1.25,
    cluster_forming_threshold=3.09,
    stat_type="Z",
    two_sided=True,
)

Examples

from geo_infer_spm.models.data_models import SPMData
import numpy as np
data = SPMData(
    data=np.random.randn(100, 50),
    coordinates=np.column_stack([
        np.random.uniform(-90, 90, 100),   # latitudes
        np.random.uniform(-180, 180, 100)   # longitudes
    ])
)

Agent interfaces

Use uv run --frozen geo-infer catalog SPM to discover public source APIs, then geo-infer describe before calling a selected target. Scripts use geo_infer_api.agent_runtime.AgentSession; the optional stdio MCP server is geo-infer-mcp. See requests, native objects and composition. The adapter retains this module's dependencies and authorization requirements.

Guidelines

• Coordinates must be valid: latitude ∈ [-90, 90], longitude ∈ [-180, 180]
• GLM implementation is Alpha status — spatial design matrices in progress
• Time series explorer uses Plotly for interactive mean±SD visualization
• Test from the workspace root:

uv run python GEO-INFER-TEST/run_unified_tests.py --module SPM

• Native Bayesian fitting uses two chains and one sampling process by default;

set chains and cores explicitly when allocating parallel CPU work.

• On macOS 27 with Apple Clang, use the explicit native compiler profile in

native sampling profiles. Linux uses the standard compiler profile; both profiles run actual PyMC and CLinker tests.

Integrations

• None required: SPM is fully self-contained (no geo_infer_* imports in src/)

and consumes plain numpy arrays / GeoDataFrames.

• Optional extras: bayesian (pymc + arviz enable the full MCMC path in

BayesianSPM; without them it falls back to a logged empirical-Bayes approximation).