Applies to:  @RISK 6.x and Up

Issue:

Inputs don't appear on the Tornado Diagram

Cause:

This behavior is usually just normal statistical variability.

The Change in Output Statistic tornado does not measure how an input influences the output.

It simply compares two statistical sub-samples:

• One bar end: the mean of output values from the lowest 10% of input values

• Other bar end: the mean of output values from the highest 10% of input values

Even if an input has no real effect, the means of two random sub-samples will differ slightly.

Check the standard error of the output (SD ÷ √iterations). If the bar size is within a couple of standard errors, the difference is just random noise.
If it exceeds that, it may be statistically significant.

Keep in mind: this graph does not show input impact. For impact, use the regression coefficients tornado. Even a low-impact input can show a difference here if it’s correlated with other inputs.

Related Articles:

Interpreting Change in Output Statistic in Tornado Graphs

Interpreting Regression Coefficients in Tornado Graphs

 All Articles about Tornado Charts