Top 2-way interaction strength tables for Fit Model and Discover Key Predictors with TreeNet® Regression

Note

This command is available with the Predictive Analytics Module. Click here for more information about how to activate the module.

Find interpretation guidance for the interaction strength tables.

The top 2-way interaction strength tables identify the pairs of variables that have the strongest interactions. The interaction tables display the % of total squared error and/or the % of squared error for the strongest 2-way interactions. Use the % of total squared error to describe the strength of the interaction relative to the variation in the data. Use the % of squared error for the specific pair of variables to describe the strength of the interaction relative to the strength of the main effects of the variables.

Interactions are not possible when a tree has only 2 terminal nodes. Thus, the maximum terminal nodes per tree must be 3 or larger. You can set this on the Options subdialog.
Note

Minitab does not display the interaction tables if all interactions have a % of total squared error or a % of squared error less than 10%.

Interpretation

In this example, the eight strongest 2-way interactions are the same for both tables; however, the ordering varies slightly. In the first table, the interaction between Annual Income and Front End Ratio is the strongest 2-way interaction. The percent of total squared error is 11.71977 which means that 11.71977% of the total squared error is explained by the main effects of Annual Income and Front End Ratio and their 2-way interaction effect.

For the same 2-way interaction between Annual Income and Front End Ratio, the percent squared error for the predictor pair with main and interaction effects is 19.73464%.

To calculate, 19.73464% = Component 3 / ( Component 1 + Component 2 + Component 3) * 100%
  • Component 1 = the squared error explained by the first main effect, Annual Income
  • Component 2 = the squared error explained by the second main effect, Front End Ratio
  • Component 3 = the squared error explained by the interaction between Annual Income and Front End Ratio and their main effects

TreeNet® Regression: Loan Amount vs Annual Incom, Income Ratio, ...

Top 2-Way Interaction Strength % of Total Squared Error Predictor 1 Predictor 2 11.71977 Annual Income Front End Ratio 9.26333 County Code Core Based Statisti 7.78507 Core Based Statisti Annual Income 5.63338 Income Ratio Front End Ratio 4.36461 Number of Units Core Based Statisti 1.26633 County Code Co-Borrower Age 1.14108 Occupancy Code County Code 1.13207 County Code Tract Income Strength: Percent of total squared error explained by a 2-way interaction
% of Squared Error Predictor 1 Predictor 2 25.21549 County Code Core Based Statisti 19.73464 Annual Income Front End Ratio 15.29069 County Code Co-Borrower Age 14.88112 Occupancy Code County Code 13.80494 Income Ratio Front End Ratio 13.39658 County Code Tract Income 11.60658 Number of Units Core Based Statisti 10.95376 Core Based Statisti Annual Income Squared Error: Specific to each predictor pair with main and interaction effects Strength: Percent of the squared error explained by the pair interaction
By using this site you agree to the use of cookies for analytics and personalized content.  Read our policy