Turnbull Results

After you perform a Turnbull analysis, you will see three new datasets in the data view, a cluster set in the spatial weights and cluster sets view, and a new map.

The new datasets are: the test statistic, p-values, and a significance dataset. The test statistic is the number of cases in the window centered on that location. The p-value is the Monte Carlo p-value, which comes from comparing the observed test statistic to the reference distribution created by randomizing the cases across the study area. The significance dataset filters the results into several categories: missing (locations with no data), non-significant, significant (at the alpha level specified in the advanced parameters), and "pop > threshold". When the population in the location is greater than the population threshold, it cannot be evaluated for clustering. In essence, this is a signal you need to raise your population threshold and re-test the data.

Map and cluster set

The map shows the significance dataset. You can also use the view spatial weights option to see the cluster set created by your Turnbull analysis.  Right click on the map to reveal the "Show object neighbors" option, and then choose the appropriate geography (States in the example below).  The choice of weight sets will now include your Turnbull cluster set, which you can select by clicking on the name (see below).

The cluster set for each location is the set of points or polygons that were required to meet the population threshold for the ego location. The weights for each of the locations in the set will be listed in the Log View as you click on them in the map.

• For point geographies, the weights will be 1 except for the last location, which may have a fractional weight.

• For polygon geographies, the weight of the ego location will be 1 if it is a significant cluster (locations that have populations greater than the threshold will have fractional weights, they need to be discounted to meet the threshold. The included neighbors will also be listed with their weight.

You can turn off the cluster set by choosing "Show object neighbors" and then clearing the checkbox next to the cluster set's name.