| partition_column |
Same as in InputTable |
[Column appears once for each specified partition_column.] Defines a partition of VIF scores. |
| attribute |
VARCHAR |
Name of target_column. |
| vif_score |
DOUBLE PRECISION |
Variance inflation factor (VIF) score of attribute at specified iteration number in iteration column.The function generates a VIF score of NaN if either of the following conditions is true: - The correlation matrix is singular.
- The variance of any attribute is 0, which happens if there is no data or each row has the same value for the entire attribute.
|
| iteration |
INTEGER |
Number of VIF iteration. Meaning depends on OutputSummary:
| OutputSummary |
Iteration Number |
|---|
| 'true' |
Number of VIF iteration at which attribute was identified as multicollinear based on vif_score exceedingvif_threshold. If vif_score never exceeded vif_threshold, number of final VIF iteration.
|
| 'false' |
Number of sequentially increasing VIF iteration. |
|
| multicollinear |
VARCHAR |
Whether attribute has vif_score greater than vif_threshold ('yes' or 'no'). Meaning depends on OutputSummary:
| OutputSummary |
Iteration Number |
|---|
| 'true' |
Whether attribute is multicollinear. |
| 'false' |
Whether vif_score exceeded vif_threshold in the iteration. Multiple attributes can have value 'yes'. At each iteration, attribute with highest vif_score above vif_threshold is removed from data set for next iteration. Attribute with vif_score higher than vif_threshold in one iteration may have lower vif_score in another iteration, after removal of another strongly multicollinear attribute.
|
This value is never marked 'yes' (multicollinear) for exception_attribute, even if its vif_score exceeds vif_threshold.
|