|
Systematic Sensitivity Analysis It is now easy (though may be time consuming) to carry out Systematic Sensitivity Analysis (SSA) with respect to variations in parameters. The methods and the way results are reported are similar to SSA with RunGEM or RunGTAP. When you carry out a dynamic simulation (over several periods), your main results are probably the cumulative differences between the Policy and the Base or Base Rerun. You may wish to know how sensitive these results are to changes in the values of the parameters used. That is, by how much would your results change if the parameters had different values. SSA with respect to parameters can help you answer this question. You specify which parameters you wish to vary and by how much and according to which distribution. Then the software •selects strategically chosen values of the parameters (different from the original ones but consistent with how you specified the parameters are to vary), •carries out the whole dynamic simulation (Base, Rerun, Policy) for all periods with these different chosen values of the parameters, •works out the mean and standard deviation of all results (eg, all cumulative differences after any number of periods). •allows you to see these means and standard deviations. When you wish to see how sensitive your results are to variations in certain parameters, the software will need to carry out the whole dynamic simulation several times. How many times depends on how many parameters are varying and whether or not they are varying independently. The number of times needed can be as little as 2 or as many as 100 or more. Hence the remark above that this may be time consuming. RunDynam can carry out several SSA solves concurrently and this will reduce the time taken if you have a computer with several processors. If you have had experience with SSA using RunGEM or RunGTAP, you will find it easy to adapt to RunDynam. The information you tell the SSA procedure and the way the software asks for this is essentially identical. •The number of solves is the same. The software uses a Gaussian quadrature to choose the values of the parameters for which the model is solved. With RunDynam, each solve is a whole dynamic simulation. •The results are presented in a similar way. But instead of a single Solution file containing say means and standard deviations of cumulative differences, you get a pair of such Solution files for each period in the Policy run. In this Help file, we concentrate on the RunDynam-specific features of SSA. We strongly advise you to read the material in the SSA Help file (called SSA.CHM) supplied with RunGEM and RunGTAP. This file is also supplied with RunDynam. Note that Sensitivity Information Files (.SIN files) can be used when doing SSA wrt parameters, just as with RunGEM or RunGTAP. You can use these files to indicate that one parameter must vary with another. As with RunGEM and RunGTAP, the ,SIN file must have the same file name as your model EXE to be relevant. [If your model EXE is USAGE.EXE, then the SIN file must be called USAGE.SIN.] At present you can only do SSA wrt parameters under RunDynam. SSA wrt shocks is not implemented (though may be at some distant date in the future). Please also note that there are some restrictions on when you can do SSA. Please see the following topics: SSA Methods with RunGEM and RunGTAP Keeping separate SSA parameter files URL of this topic: www.copsmodels.com/webhelp/rundynam/hc_ssa.htm Link to full GEMPACK Manual Link to GEMPACK homepage |