Local empirical bayes
Empirical Bayes estimators consider that the "real" rate associated with each area is not known, and that observed rate is available. The idea of Bayesian estimators is assumed that the actual rate is a random variable having a mean and a known variance.
The best estimator in this rate is a linear combination of the observed rate (events / population) in area A and an average value B weighted by a factor C.
The average weight may be used in the average rate of the surrounding areas, in this case, the method is called Local Empirical Bayes.
It is accessible through:
PROCESSING → SPATIAL ANALYSIS → LOCAL EMPIRICAL BAYES
Input Information:
Layer Name: Select the input layer.
Load GPM: Loads a gpm from file, if not, creates a new one.
Attribute Link: Defines the attribute that identifies the objects of this layer.
GPM: Opens a dialog to select a file with a desired proximity matrix.
Parameters:
Event Attr Name: Attribute name that identifies the desired event information.
Population Attr Name: Attribute name that identifies the population information.
Rate Correction: This is the multiplicative rate correction value.
Output Information:
Repository:
Select one type
of repository by clicking on the
button to save the output layer as a file or on the
button to save it in the database.
Layer Name: Defines the name to create the output layer.
Click on the OK button and then the local bayes rate will be calculated.
Note: To calculate Bayesian Rates, the denominator (population at risk) does not have values equal or smaller than zero.