 |
Download a pdf copy of this help file here |
|
Download a pdf copy of this help file here |
Tornado plots describe how sensitive the value of an output variable is to the input variables of the model.
Tornado plots can be produced in ModelRisk by selecting an output variable in the Simulation Results window and clicking:
Producing a tornado plot requires making the following choices:
1. Select the output of interest
2. Select the statistic of interest by clicking on one of the following icons:
Rank
correlation (This is the most common analysis used in Monte Carlo models)
Proportional
contribution to variance (The fraction of the output variance attributable
to each input, where negative values reflect negative correlation)
Contribution
to variance (The amount of output variance attributable to each input,
where negative values reflect negative correlation)
Output
conditional mean (requires tranches)
Output
conditional standard deviation (requires tranches)
Output
conditional coefficient of variation (requires tranches)
Output
conditional percentile (requires tranches, and percentile)
Tranches are used to organise the simulation data into equal groups for a specific input variable. For example, if 20 tranches are specified, ModelRisk divides the simulation data into 20 equal groups that correspond to the 0-5%, 5-10%, …, 95-100% ranked data for an individual input. It then determines the output statistic (like the mean) for each of these sub-sets of the simulation data, and plots the minimum and maximum of the output statistic across all these sub-sets in a tornado chart. This shows how much the output statistic can vary depending on what the value of the input variable might be. As a general rule, the more samples of the model you run in a simulation the greater the number of tranches you can use and the more precise the tornado chart will become.
If Percentile has been selected, click 
to
define the required percentile
If the statistical analysis requires tranches, select the number of
tranches to be used by clicking 
. The number of tranches define the number of points that will be tested
each input variable, similar to the spider plot. The graph then plots
from the lowest to the highest values to create each tornado bar.