User manual PALISADE RISK 5.5

[. . . ] Guide to Using @RISK Risk Analysis and Simulation Add-In for Microsoft Excel ® Version 5. 5 May, 2009 Palisade Corporation 798 Cascadilla St. Ithaca, NY USA 14850 (607) 277-8000 (607) 277-8001 (fax) http://www. palisade. com (website) sales@palisade. com (e-mail) Copyright Notice Copyright © 2009, Palisade Corporation. Trademark Acknowledgments Microsoft, Excel and Windows are registered trademarks of Microsoft, Inc. IBM is a registered trademark of International Business Machines, Inc. Palisade, TopRank, BestFit and RISKview are registered trademarks of Palisade Corporation. [. . . ] Smart Sensitivity Analysis is done because it is possible for simulation data to show a correlation between an input and an output when, in reality, the input has no effect on the output in your model. Without Smart Sensitivity Analysis, tornado graph bars may be displayed for those unrelated inputs. Reference: @RISK Commands 373 There are isolated instances where you should disable Smart Sensitivity Analysis on the Sampling tab of the Simulation Settings dialog box to improve performance and sensitivity analysis results: 1) The Smart Sensitivity Analysis setup time for scanning precedents at the start of the simulation adds significantly to runtime of a very large model and you are not concerned that you may see sensitivity analysis results (or tornado graph bars) for inputs unrelated to outputs. 2) You use a macro or DLL that performs calculations using @RISK input values in cells that have no relationship via workbook formulas with the output. This macro or DLL then returns a result to a cell that is used in calculating the outputs value. In this case there is no relationship in workbook formulas between the output and the @RISK distributions and Smart Sensitivity Analysis should be disabled. To avoid situations like this, we recommend that you create macro functions (UDFs) that explicitly reference all used input cells in their argument lists. In earlier versions of @RISK, Smart Sensitivity Analysis was not used. This is the equivalent to the Settings menu Simulation Settings command Smart Sensitivity Analysis Disabled option. Regression and Correlation Two methods -- Multivariate Stepwise Regression and Rank Order Correlation -- are used for calculating sensitivity analysis results, as discussed here. Regression is simply another term for fitting data to a theoretical equation. You may have heard of the "Least-Squares" method, which is a type of linear regression. Multiple regression tries to fit multiple input data sets to a planar equation that could produce the output data set. The sensitivity values, returned by @RISK, are normalized variations of the regression coefficients. Stepwise regression is a technique for calculating regression values with multiple input values. Other techniques exist for calculating multiple regressions, but the stepwise regression technique is preferable for large numbers of inputs, since it removes all variables that provide an insignificant contribution from the model. The coefficients listed in the @RISK sensitivity report are normalized regression coefficients associated with each input. A regression value of 0 indicates that there is no significant relationship between the input and the output, while a regression value of 1 or -1 indicates a 1 374 Results Commands or -1 standard deviation change in the output for a 1 standard deviation change in the input. The R-squared value, listed at the top of the column, is simply a measurement of the percentage of variation that is explained by the linear relationship. If this number is less than ~ 60% then the linear regression does not sufficiently explain the relationship between the inputs and outputs, and another method of analysis should be used. Even if your sensitivity analysis produces a relationship with a large value of R-squared, examine the results to verify that they are reasonable. Mapped values are simply a transformation of the beta coefficient for the Regression (Coefficient) into actual values. The beta coefficient indicates the number of standard deviations the output will change, given a one standard deviation change in the input (assuming all other variables are held constant). [. . . ] 191, 286 Chi-Squared. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190, 286 Detailed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367 Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 Kolmogorov-Smirnov (K-S). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [. . . ]