Even though the data is multi-dimensional, it is often desirable to see one dimensional graphs or contour plots. Doing this causes several side effects that we will discuss now. First make a 1-d plot of the a2.mbl data.
Model -> Graph -> Dimensions 1
Graph
This graph plots the first input attribute against the output attribute while ignoring the second input attribute. Thus, it is now possible to see the output values of the data points. At first glance the fit to the data seems incorrect, but that is not the case. All of the data is presented in the graph regardless of what the value of its second input attribute is. The curve, however, shows the approximated function with the second input attribute held constant at its value in the query point. At the bottom of the output, the query is shown to be the point [5, 5]. If we back up and look at the page with the contour plot, we see that when the second attribute is held at a value of 5, the approximated function goes through the middle of the valley at the center and between the peaks at each side. This explains the 1-d graph with the curve near the bottom of the data. It is a graph of how the approximated function varies with the first input attribute while the second is held fixed. If we change the value of the second input attribute to 8 and repeat the graph, the curve will move near the top of the data.
Edit -> Query -> ``x2'' 8 Model -> Graph -> Graph
It is also possible to see how the function looks when the first input attribute is held fixed and the second is varied.
Model -> Graph -> x attribute -> ``x2''
Graph
This graph looks very similar to the original graph using the ``x1'' input attribute. That is because the data was generated from a function symmetric with respect to the inputs. Finally, it is possible to generate multiple graphs at one time. Producing graphs for each input is done as follows:
Model -> Graph -> x attribute -> ``[All inputs]''
Graph
When the data has more than two inputs, it may be desirable to graph contour plots on pairs of the inputs, while leaving the other inputs fixed at their query point values. This can be done in a way similar to the 1-d graphs.