## Overview

This is a python module for calculating global (Moran's I [1]) and local spatial autocorrelation [1.5] using the AMOEBA algorithm [2]. This code works on shapefiles, although a base class is provided to allow the examination of other objects, e.g. from a spatial database.

## Usage

The easiest way is to call autocorrelate.py with the name and path of the shapefile, e.g.:

python autocorrelate.py path/to/file/filename.shp

To use in other python programs:

from lcia_autocorrelation.ac_shapefile import AutocorrelationShapefile ac = AutocorrelationShapefile("filepath") ac.global_autocorrelation()

Autocorrelation calculations are made using the PySAL library; multiple measures of autocorrelation are possible.

## Local Indicators of Spatial Autocorrelation (LISA)

Moran's I (http://en.wikipedia.org/wiki/Moran's_I) is a single statistic for glboal autocorrelation. However, the calculation of Moran's I involves summing the individual cross products of each spatial unit. Local indicators of spatial association (LISA) (Anselin, L. (1995). "Local indicators of spatial association – LISA". Geographical Analysis, 27, 93-115) uses these local indicators directly, to calculate a local measure of clustering or autocorrelation. The LISA statistic is:

Where *I* is the autocorrelation statistic, *Z* is the deviation of the variable of interest from the average, and *W* is the spatial weight linking **i** to **j**.

We use the PySAL library (http://code.google.com/p/pysal/) to calculate LISA statistics (http://pysal.org/users/tutorials/autocorrelation.html#local-indicators-of-spatial-association).

## Installation

Using pip:

pip install lcia-autocorrelation

Using easy_install:

easy_install lcia-autocorrelation

### Requirements

The following packages are required

- numpy
- scipy
- pysal
- rtree
- osgeo
- django
- progressbar

## Copyright and License

This code was written by Chris Mutel [3] during his studies at ETH Zurich [4], and is copyright 2011 ETH Zurich. The license is 2-clause BSD.