Spatial autocorrelation
why spatial autocorrelation
Spatial autocorrelation is the correlation between values of a variable in a geographic space. It is important for several reasons:
-
Pattern identification: It helps identify spatial clusters, hotspots, or outliers, providing insights into phenomena like crime rates or disease outbreaks.
-
Data exploration: It allows researchers to understand how variables are related in space, such as examining the relationship between income levels and educational attainment.
-
Spatial modeling: Accounting for spatial autocorrelation in models improves accuracy and reliability of estimates, avoiding biased results and invalid inferences.
-
Spatial planning and policy-making: Understanding spatial autocorrelation aids in identifying areas of need or suitable locations for infrastructure development and allocation of public services.
-
Geostatistics and spatial prediction: Spatial autocorrelation is used in geostatistics to predict values at unobserved locations and create spatially continuous maps.
Overall, studying spatial autocorrelation helps us understand spatial processes, detect patterns, improve analyses, and inform decision-making in various disciplines dealing with spatial data.
types of autocorrelation
There are 2 types of auto correlation:
1.Global which means by a single value which applies to the entire dataset
- study the entire map pattern
- generally asks the quetion as to whether the pattern displays clustering or not
- decide whether or not an overall configuration is autocorrelated, but not where the unusual interactions are.
2. Local which a value calculated for each observation unit
- focus on exploring within the global pattern to identify individual clusters that maybe either driving the overall map pattern or displaying a deviation from the pattern.
Joint-count statistics
- provide a simple way of quantitatively measuring spatial autocorrelation for a set of spatially adjacent polygons.
- applicable only to nominal data or other measurement level converted to nominal.
Morans I Autocorrelation
- interval or ratio data type
- output of Morans I spatial autocorrelation coefficient:
- -1, Negative autocorrelation / Disperse
- 0, No autocorrelation / Random
- +1, Positive autocorrelation/ Cluster
Geary's C Spatial Autocorrelation
- the 2 neighbouring values arev compared with each other directly
- Interpretation of Geary's C
- C of approximately 1 indicates no autocorrelation/ random
- C of 0 indicates perfect positive autocorrelation/ clustered
- C of 2 indicates perfect negative autocorrelation/ dispersed