Applies to: @RISK 5.x and Up

 

Choosing a Sensitivity Measure in @RISK

After running a simulation, the Sensitivity Analysis window provides an overview of how inputs affect outputs. Access it from the Results section of the @RISK ribbon by clicking the small tornado icon. You can also click the tornado at the bottom of the Browse Results window for a specific output.

 

Change in Output Statistic

• This method compares differences in output statistics (mean, mode, or a specific percentile) between extreme bins.

• Click the % icon at the bottom of the Sensitivity Analysis window or the tornado icon in Browse Results to select the statistic.

• The Change in Output Statistic tornado highlights differences at the extremes, while the spider graph shows both direction and differences across all bins.

 

Regression or Correlation Coefficients

• Regression coefficients and regression mapped values are scaled versions of each other.

• Correlation coefficients are rank-order correlations, effective for linear or non-linear relationships.

• In the Sensitivity Analysis window, select Display Significant Inputs Using: Regression (Coefficients) to see R² values.

  • Low R²: Linear regression is a poor fit; focus on correlation coefficients.

  • High R²: Regression fits well, but check for multicollinearity. Discrepancies between regression and rank-order correlation may indicate problems.

 

Contribution to Variance

• R² measures how much variance in an output can be explained by the inputs rather than noise or sampling error.

• Variables are added one by one to the regression, and the increase in R² shows their contribution to variance.

• Unlike regression coefficients, this measure is independent of input magnitude and focuses on predictive improvement.