Centroid or Distance Summarization Method
Use this feature to specify the summarization method.
| • | When the dependent variable is nominal and you select Class Centroids in the previous option (For Nominal Dependent Variables, compute distances to:), this is the method for computing class centroids in the training set. |
| • | When the dependent variable is nominal and you select Each Observation in the previous option (For Nominal Dependent Variables, compute distances to:), this is the method for summarizing the individual distances within a class. |
| • | When the dependent variable is continuous, this is the method for computing the predicted value for a particular observation from all individual distances, using observed values of the dependent variable from the training set and inverse distances as weights. |
Options for summarizing centroid/distance are listed in the following table:
|
Method |
Description |
|
MEAN |
Computes the centroid distance between the mean of each group. |
|
MEDIAN |
Computes the centroid distance between the median of each group. |
|
MIN |
Computes the centroid distance between the minimal values of each group. |
|
P1 |
|
|
P5 |
1 |
|
P10 |
1 |
|
P25 |
1 |
|
P75 |
1 |
|
P90 |
1 |
|
P95 |
1 |
|
P99 |
1 |
|
MAX |
Computes the centroid distance between the maximal values of each group. |
To Specify the Method:
| 8 | Select the method using the drop-down menu. |