Data Selection Wizard :: Data (Model-Based Calibration Toolbox™)

Use the first screen of the wizard to select the data set to build models from. You can also select whether to use all the data set or to match the data to a design, if any designs are in use in the test plan. Designs appear in the left list box, data sets on the right, and you can click the button to load a new data set.

Step 2: Select Input Signals

Select the input signals for the model (on all stages of the hierarchical model) from the list box of data signals on the right, and match them to the correct model variables using the big button with the arrow. Double-clicking an item in the data signals list selects that signal for the currently selected input factor (with the range if the Copy range check box is selected) and then moves to the next input.

If you entered the correct signal name at the model setup stage, the appropriate signal is automatically selected as each model input factor is selected. This can be time-saving if there are many data signals in a data set. If the signal name is not correct, you must click to select the correct variable for each input.

Select the check box Copy range if you want to use the range of the selected data signal for the model input range. Ranges are not automatically copied, although stored templates have the ranges that were set when the template was saved.

If you want to exit without selecting response models, you can click Finish. For example, you may want to create boundary models or designs without building response models. If you do not want to inspect or edit the data, clear the check box Open Data Editor on completion.

To continue to selecting response models, click Next.

Step 3: Select Response Models

Use the following controls to set up your models:

If starting with an empty Responses list box, select the desired response in the Unused data signals list, and click Add.
If you are using a test plan template, responses may already be specified in the Responses list box. If you want to change a response, select a signal and click the large button with the arrow to replace the current selected response. The previous response appears in brackets to inform you what has changed.
When there is already a response in the list box, clicking Add does not replace the selection but increases the number of responses. The replace button (with the arrow) is not available when the Responses box is empty.
You can use Delete to remove selected responses.
You can select Datum models (if the local model supports them), and you can use the Set Up buttons to change the local and global models. See Local Model Setup, Global Model Setup, and Datum Models for details.

If the check box Open Data Editor on completion is selected, when you click Finish the Data Editor opens to select data for modeling. The models, however, are not built until you close the Data Editor. Clear the check box Open Data Editor on completion if you have already inspected the data, and want to use all the data for modeling. When you click Finish the models are built.

You can only click Next if you are matching data to a design. See Step 4: Set Tolerances.

Step 4: Set Tolerances

You only reach Step 4 if you are matching data to a design. Setting tolerances is only relevant if you are matching data to a design. You can only match data to designs for global models.

You can also edit tolerances later using the Tools menu in the Data Editor window. See The Tolerance Editor for definitions of clusters.

Set Input factor tolerances for each variable to determine the size of the tolerance in each dimension. This is used for selecting data. The tolerance in each dimension determines the size of "clusters" centered on each design point. Data points that lie within tolerance of a design point are included in that cluster. Data points that fall inside the tolerance of more than one design point form a single cluster containing all those design and data points. If no data points fall within tolerance of a particular design point, they remain unmatched. Default tolerance values are related to variable ranges.

The choices you make in the Default data selection options determine how the cluster algorithm is first run to select matching data and design points. This only affects the status of the check boxes for data and design points when you first see the Cluster Plot view in the Data Editor. You can always alter the results of this later in the Cluster Information list view, where you can manually select the data and design points you want to use.

When you exit the Data Editor these selections determine what data is used for modeling and how design points are augmented and replaced. Selected data is used for modeling and added to the design. Data you have decided to exclude is not used for modeling or added to the design.

Unmatched data — Use or Do not use
Data that does not lie within tolerance of any design point is unmatched. You can decide what to do with these. If you select Use, this data is selected for modeling and added to your design.
If you select Do not use, then unmatched data is not used for modeling or added to the design; it is excluded data.
Clusters with more data — Use all data or Use closest match only
This refers to clusters containing more data points than design points. If you choose Use all data, all the data points in these clusters are selected for modeling and added to the design, replacing the design points in those clusters.
If you choose Use closest match only, then a one-to-one match of the data point closest to each design point is selected, and these are the only points that are selected for modeling and added to the design (replacing a design point each).
Cluster with more design — Do not replace design points or Replace design with closest
All data from these clusters is selected for modeling. The setting here only affects selections for the design.
Where clusters contain more design points than data points, you can choose to leave the design unchanged by selecting Do not replace design points.
If you choose Replace design with closest, this replaces the design points where possible with the closest data point and leaves the rest of the design points unchanged.

Remember you can override any of these selections manually in the Data Editor; changes are only applied when you close the Data Editor after matching. The cluster plots view in the Data Editor window appears by default when you close the Data Wizard while you are matching data to designs. See the next section, Matching Data to Designs.

Test Groupings

Matching Data to Designs

Recommended Products

Get MATLAB trial software

Includes the most popular MATLAB recorded presentations with Q&A sessions led by MATLAB experts.

Get the Interactive Kit

Search MATLAB Documentation
R2012a Documentation → Model-Based Calibration Toolbox
View documentation for other releases	Learn more about Model-Based Calibration Toolbox