Graph Tab

The Graph tab of the Analyze Model window shows a condition versus age graph, which contains both a curve and data points:

• The curve results from the mathematical equation underlying the model. In a sense, it is the ideal curve for this particular model.

• The data points are derived from regression analysis for just the selected condition attribute and node (and child nodes, if any). (This may or may not be the entire data set.) The data points can be colored black or gray, with the black data points denoting outliers.

  Note: Outliers are only plotted if the model selected in the Models tab is the one that has the 0-Selected value in the Model Run column.

When you hold your mouse cursor over a point or line within the graph, the application displays a popup window showing the Index and Age values at that point.

When you right-click the tab, the system displays a shortcut menu with the following special command:

• Edit Outliers: This command is only available when the Graph tab shows the graph of one model and that model is the model designated 0-Selected in the Model Run column of the Models tab. (The model usually identified as this is the one assigned to the node selected in the Performance Models window or, if none is assigned to the node, the model selected in the Models pane of the Performance Models window.)

When the edit outliers command is available, you can use it to perform outlier analysis. User-defined outlier criteria are a function of condition attribute value and age. Multiple outlier criteria can be set.

1. Select Edit Outliers, and the User-Defined Outliers window opens. 
2. Right-click in the window and choose Insert. The outlier criteria are additive. The example below shows a portion of this window where young, poor data points (those with an index value less than 50 or those less than five years old) and old, good data points (those with an index value greater than 85 or those older than 20) are defined as outliers:
   
3. Once the outlier definitions are as desired, click OK to close the User-Defined Outliers window.
4. You can then include the new outlier definitions in the model selected for detailed analysis by clicking the Models tab. In the Models tab, the system will display a message to advise that the outlier definitions have changed.
5. Click Yes to continue, and then the new outlier definitions will be saved and used in displaying data points on the Graph tab.

Models Tab

The Models tab of the Analyze Models window is where you perform all regression analysis (including the selection of what model form to use in the regression analysis) and where you identify which models to graph in the Graph tab.
The tab provides two panes (Current Models on the left and Model Development on the right) along with the Model Type field. This field identifies the model form whose models will be shown in the Current Models pane. The model type is selected from a drop-down list, which currently has seven types. See Model Equations and Regression Methods for more information on the different model types.

Current Models Pane

The Current Models pane shows all models of the type selected in the Model Type field. (The list of models is taken from the Models window.)

When you right-click a model in the table, the system displays a shortcut menu with the following special command:

• Move Model: This command moves the selected model into the Model Development pane, where it can be graphed against the current set of data in the Graph tab and where it becomes a candidate to be the Default Model Structure model for the currently selected node.

Model Development Pane

The Model Development pane on the right side of the tab is where you perform regression analysis as well as where you identify which models to graph in the Graph tab.

Each record in the Model Development pane represents a particular model, regression method, and filter on the data set. When this tab is first displayed, only one record is shown in this pane. This record contains the model for the node currently selected in the Performance Models window. (If that node did not have a model assigned, then the system uses the currently selected model in the Models pane of the Performance Models window.) This model is assigned the 0-Selected value in the Model Run column, and is the only model assigned this value.

Additional records are added to this pane by using commands found by right-clicking in this pane or by moving a model from the list of models in the left pane (by using the Move Model command found in the right-click shortcut menu).

Columns in the Model Development Pane

The main columns found in this pane are described below:

• Model Run: This column contains a drop-down list, the entries of which indicate the model's purpose or origin. The current Model Run types are described below:

0 - Selected: This value indicates the currently selected model. It initially is the model selected from the Performance Models window. Only one record can be assigned this value. The model type, or its parameters for this record, can be changed by using either the Move for Analysis or Calculate Excluding Outliers command found on the shortcut menu displayed by right-clicking the record.
1 - Regression: The Calculate Regression Models command inserts three records into the table, each of which is assigned this value. These records contain the linear regression results for the linear, exponential, and hyperbolic model types using the entire data set.
2 - Power: The Calculate Power command inserts one record into the table, and this record is assigned this value. This record contains the regression results for the power model type using the entire data set using a special "power" non-linear regression method.
3 - Existing: When a model is moved from the left pane to the right pane, the record for this model is assigned this value.
4 - Least Squares: The Find Closest Models command inserts six records into the table, each of which is assigned this value. These records contain the results of a "least squares" quasi-regression analysis for each model type using the entire data set.

• Model Type: This column identifies the type of model.
• NUM OUTL: This column is only important for the 0-Selected record (only this record will have a non-zero value in this column). It shows the number of outliers that will be excluded from analysis when you select the Calculate Excluding Outliers command.
• NUM SAMPLE: This column shows the number of data points in the data set being analyzed (that is, the number of data points pertinent to the currently selected node in the Performance Models window).
• SE: This column shows the error sum of squares of the data set (possibly with outliers excluded) against the model.
• SE COEF: This column shows the number of standard deviations outside of which data points are considered outliers. This value is used by the Define Outliers and, at times, the Calculate Excluding Outliers commands. The value is set by using the Set Standard Error command.
• Select: When the check box in this column is selected, the model will be displayed in the Graph tab when you click the Graph tab. Note: The Edit Outlier command found in the Graph tab is only available when the 0- Selected record is the only one selected for graphing.

Right-click Shortcut Menu Commands

When you right-click a record in this pane, the system displays a shortcut menu with the following special commands:

• Calculate Excluding Outliers: This command is only available for the model with a Model Run type of 0-Selected provided that data points exist in the data set for the model. When available, it runs a regression analysis for the model's data set with the identified outliers excluded. (Outliers are identified by either using the Define Outliers command in this pane or by using the Edit Outliers command in the Graph tab.)
• Calculate Regression Models: This command is only available for the model with a Model Run type of 0-Selected provided that data points exist in the data set for the model. When available, it performs linear regression analyses for the linear, exponential, and hyperbolic model types using the entire data set. It then creates three new records in this pane to contain the results of the analyses, with each record being assigned a Model Run type of 1-Regression.
• Define Outliers: This command is only available for the model with a Model Run type of 0-Selected provided that data points exist in the data set for the model. When available, it identifies outliers based upon the currently set Standard Error (SE) Coefficient value shown in the SE COEF column. After identifying the outliers, it then places the outlier count in the record with a Model Run type of 0-Selected. (The identified outliers are then excluded from regressions run using the Calculate Excluding Outliers command.)
• Find Closest Models: This command is available for all Model Run types, but only if data points exist in the data set for the model. It performs a "least squares" quasi-regression analysis for each model type using the entire data set. It then inserts six records in the pane, one for each model type, to contain the results of the analysis. Each record is assigned a Model Run type of 4-Least Squares.
• Move for Analysis: This command is only available for models that are not designated with a Model Run type of 0-Selected. When available, it copies the values of the record that you right-clicked into the record designated 0-Selected. Once copied, the modified 0-Selected record can then be regressed with outliers and/or added to the set of permanently available models.
• Return Current Model: This command is only available for the model with a Model Run type of 0-Selected provided the model is not already permanently available. When available, it applies the model of the 0-Selected record to the currently selected node in the Performance Models window and also includes the model in the list of permanently available models (that is, those models shown in the Setup Models window and shown in the left pane of the Models tab).
• Set Standard Error: This command is only available for the model with a Model Run type of 0-Selected provided that data points exist in the data set for the model. When available, it sets the Standard Error (SE) Coefficient value shown in the SE COEF column.