Model Exploration Tab
Select the Model exploration tab of the Response Optimization Startup Panel to access options to generate response spreadsheets and graphs of the predictive models. A response graph shows the effect on the response variable prediction of adjusting an independent variable. The independent variable can be either numeric or categorical. The response variable can also be numeric or can be categorical; in the latter case, the plot and the spreadsheets involve the confidence levels, which is continuous and can therefore be conveniently graphed.
While one specially chosen independent variable is altered (and plotted across the X-axis of the graph), values must also be provided for the other independent variables of the predictive model(s). These are given fixed values and, hence, are referred to as fixed independents. Thus, the response graph actually represents a one-dimensional slice through an N dimensional response surface, where N is the number of independent variables.
If there is more than one model present in your current analysis, the response graphs and spreadsheets also include responses for the combined models (i.e., ensemble predictions) together with their variance (for continuous dependents only). The variance measures the extent of the agreement (or disagreement) among the predictive models. Given a set of independent values, the more the models differ among each other in their predictions, the bigger the variance is.
| Element Name | Description |
|---|---|
| X-axis | In this field, select the independent variable that is to be varied across the X-axis. For classification tasks, a response spreadsheet and graph is generated for each categorical level of the response (dependent) variable |
| Y-axis | This option is available only when the response (dependent) variable is categorical (i.e., classification tasks). Select All levels from this list to print response spreadsheet/graphs for every categorical level |
| Min | Specify the minimum value of the independent variable on the X-axis. By default, this is the smallest value of the variable in the current data set |
| Max | Specify the maximum value of the independent variable on the X-axis. By default, this is the largest value of the variable in the current data set |
| Samples | Specify the number of points evaluated to generate the response spreadsheets and graphs. The samples are evenly spaced between the minimum and maximum (see above) |
| Fixed independents | Click this button to display a general user entry spreadsheet, which you can use to customize the independent variables. Click the OK button on the spreadsheet to accept the changes you have made. The Fixed independent list , which is located just below the Fixed independents button, provides you with a view of the current values of the fixed independent variables. Use the Fixed independents button to modify these values to suit your analysis needs |
| Response spreadsheet | Click this button to generate the response spreadsheets(s). If all the categorical levels of the response variable are selected (see Y-axis above), a response spreadsheet is generated for each categorical level of the response (dependent) variable. If a categorical level is specified in the Y-axis drop-down list (see above), a spreadsheet is generated only for that particular category |
| Response graph | Click this button to generate the response graph(s) for the individual models (first graph) and the ensemble (second graph). The latter is also include line plots of the ensemble errorbars should the task be regression. If all the categorical levels of the response variable are selected, a response graph is generated for each categorical level of the response (dependent) variable. If a categorical level is specified in the Y-axis drop-down list, a graph is generated only for that particular category |