If I've created a Network with a dataset and now want to use the same structure with a different dataset. Is it possible to save structure and use new data, without needing to start from scratch create variable clustering and multiple clustering, etc?
You just have to associate your new dataset to the final network (i.e. the one with the Manifests and the Factors).
Here is the workflow:
- Modeling Mode (F4);
- Data - Associate Data Source - Text File;
- The variables that are both in your network and in the new dataset will automatically take the type defined in the network (discrete or continuous). Otherwise, use the wizard to define the data type of your new variables;
- Define your missing value processing, if any;
- If some Column names are not identical to Variable names, manually associate these data columns to their corresponding variables;
- You can also add new variables to your network; in this case, you will use the last wizard to define the discretization/aggregation of these new variables.
Note that your Factors will probably not have any corresponding data in the new dataset. Thus, they will be Hidden variables, and will appear as white nodes in the network (as [Factor_6] below).
You will be able to automatically impute their values (based on their associated Manifests in the network) by selecting them and making a right click to select the Imputation Menu - Choose the Values with the Maximum Probability.
Once imputed, they will not be Hidden variables anymore, and will appear as blue nodes.