Prior to version 6.3, the following sampling methods were used:
| • | Multivariate (change all together), if selected, causes all constants to be changed together. If you are using random control to change constants, this would normally be selected. |
| • | Univariate (change one at a time), if selected, causes first one Constant to be changed and then another, with the first Constant set back to its normal simulation value (the model value or changed value entered for it). This is a useful option for doing a series of VECTOR searches across parameters. |
| • | Latin Hypercube (change together exhausting axes ranges) , if selected will cause a Latin Hypercube search to occur. A Latin Hypercube search is simply a mechanism to ensure that the full range of each parameter being varied is explored in the number of simulations specified. This is desirable for big models where each simulation takes a long time. |
| • | Latin Grid (weighted grid search), if selected, will conduct a grid search using the specified probability distributions to divide the parameter space. The number of simulations specified is the number of divisions, and will be raised to the power determined by the number of search parameters to determine the actual number of simulations required. This is similar to specifying a multivariate search with VECTOR distributions for the parameters but will sample the distributions more coarsely on the tails. |
| • | File (use a file to specify parameters), if selected, will use a file to determine what inputs to change, and ignore any constants specified in this dialog. You can use this method to simulate a sample generated by another tool or to automate a set of scenarios. This file must be a tab delimited text file (UTF8 for international characters). The first row should contain a list of model constant names (tab delimited), with the corresponding values on each successive row. If a value is missing from this file the previously used value is repeated. See Control File Format. |
You can use the pre-6.3 sampling method by setting the random seed to a negative value, which will also activate the pre-6.3 random number generator for backwards compatibility. We do no
