The previous method to specify Predefined criteria was limited to a maximum of 6 ensembles as regulated by the switch PreDefinedEnsembles in the technical module. From current version the swith will have only 2 modes off/on and instead the number of criteria will be decided from the information in the parameter table Criteria for Ensembles. The second column in table named Ensemble can have any sequence of number and the number of ensemble created will correspond to the highest number given among all rows in the table. Please note that the Number of Elements that corresponds to the number of rows in the table does not have a direct link to the number of ensembles except that the number given in the Ensemble column should be less or equal to the Number of Elements. Note the example below where each new row in the table corresponds to a new ensemble.
In the last column a new parameter is introduced namely MR Dim1 Index. This parameter will te used as a filter to select only those candidates that correponds to the value of the index in Dimension 1 of the calibration as specificied in the MultiRun tab
For the example above a sequence of runs will have Index for 1 to 5. Providing the MR dim1 index is assigned to 0 instead of a value from 1 to 5 all candidates will be considered in the statistic for that Ensemble. By selected 1 to 5 only those specific candidates that wil have a match with the index will be considered. Also note the Filter Var Value in the table that corresponds to the specific treatment that will be used based on the value in the validation file (see Version 6.2.6 New method to handle Multiple Treatment or Replicates.
Once a MultiRun with predefined criteria has been completed a table with the complete results of the statistic should be ready in the Evaluation Multi windows. In the example above 7 ensembles should be included as can be views the Ensemble – Split tab.
The Performance indicators describe which method have been used to transform and filter the candidates. Normal Corresponds to no transformation (0- Transformation Table). A mean value of all performance indicators corresponds to transformation 4. The implication of this is that this ensemble will represent the best possible performance of all treatments as selected by the filter variable. This ensemble (2) should be most interesting if you are searching for parameter candidates that make the best common performance using the same parameter values for all treatments. Proving that the you are interested to consider each treatment independent of the other you should select the ensemble 3-7 for you interpretations. The possible difference in parameter values between the treatments will be easy to identify.
To View about the exact criteria that have been used to select the 50 acceptable candidates you can find those in the tab Defined Criterias
The view of the accepted run for all the 7 ensemble will be be voew as a time serie if you click the corresponding box if the Validation tab
as below with run no on x-axis and all candidates compared to accepted candidates
Or as in the exemple as the cumulative distribution curve as shown below also for parameters
The parameters corresponds to the Dim 2 parameters that are visible in the MultiRun tab
The data for the example above is available from CoupModel Samples with link for all the data. FlakaLidenForestFertilizationTrial. The data has been put together for demonstration purpose by Harald Grip and myself. A short video will be prepared to show how the data are organized for this model application.