A Value Tornado Diagram is a type of sensitivity analysis that is useful when you want to see the relative impacts on the objective function value and the optimal policy when the values in your model are varied from a specified low setting to a specified high setting. A Value Tornado helps you in deciding which of several values to model probabilistically. It is also a very useful tool for presenting your work to others.

In a Value Tornado, each bar represents a sensitivity analysis on a single variable. Only variables initialized with a constant expression in your model can be analyzed in a Value Tornado.

__Results and Policy Changes__

To produce a Value Tornado Diagram for a model, DPL runs the model several times. First, DPL runs the model once to establish the base results. This base results are displayed as a vertical line. Then, it runs the model twice for each variable in the Value Tornado: once using the specified low value for the variable and once using the specified high value. The width of the bar for the value is the difference in the expected value of the objective function between the two runs. The expected value of the original model is indicated by a vertical line. In each of these runs, it is a complete run of the model. I.e., if the model is probabilistic (has chance nodes), DPL does a complete decision analysis and rolls-back the tree to determine the expected objective function value in each run.

In most cases, the output values for the low value and the high value will not both be less than or not both be greater than the base results, so the ends of the bar will lie on either side of the vertical line for the base results. In cases where both output values are less than or both are greater than the base results, the bar ends for low and high both lie on the same side of the vertical line. DPL draws the bar in this case with two half-height bar segments on the same side of the vertical base results line.

The bars indicate changes in the optimal policy through color changes. The color change occurs at the midpoint between the line indicating the base results and the end of the bar. The position of the color change does not indicate the precise point where the policy change occurs. Green indicates the base result optimal policy; blue indicates that the low value for the variable causes a policy change; and magenta indicates that the high value for the variable causes a policy change.

__Adding Additional Variables__

To add additional bars to the diagram, choose Chart | Series | Data | Modify. As you add more sensitivity bars to the diagram, DPL will automatically sort them so the widest bar appears at the top and the narrowest bar appears at the bottom. If there are too many sensitivity bars to display on the screen at once, a scroll bar at the right of the diagram allows you to scroll to view the sensitivity bars.

__Ranges for Sensitivity Analysis__

The high and low values you choose for the sensitivity variable are up to you. It is important that the ranges you specify for each variable in the comparison represent roughly the same degree of uncertainty (for example, encompass 80% of the total distribution) so the comparison will be meaningful. You may get unexpected results if you enter high and low values that do not span the nominal value. To change the range for a variable you have already run, go to Chart | Series | Data | Modify and modify the low or high settings for the variable.

__Updating the Tornado Diagram__

If you change your model and wish to update the bars in the diagram, choose Chart | Series | Data | Refresh. DPL will re-run the sensitivity analysis using the current model and previously defined variables and ranges. If a variable in the Tornado diagram is no longer in the model, DPL will warn you and if you choose to update the diagram the variable will be removed.

* Versions:* DPL Professional, DPL Enterprise, DPL Portfolio

*See Also*