does anybody know, if it is possible to create a Response surface mesh and 3D contoour plot in JMP. If it is, how can I do that? I added two pictures, to better ilustrate what I have in mind.
Response surface mesh
3D contour plot
Is there any better way to present a response which is dependent on three factors in one plot? I first presented the relation with 9 surface plots (3 for every combination of 2 factors, the third factor has always a different value (-1,0,1)), which is a very clear presentation in my opinion but it takes a lot of space and can hide the general picture because it is too segmented. I was also thinking about cube plot, but it hides a lot of information, especially, because I have one quadratic effect...
Thank you in advance,
You will find a lot of information about the JMP versions by selecting Help > Books > Profilers.
I should add that these profilers are available in most fitting platforms and separately as their own stand alone platform. In the latter case, they use a column formula. You can save fitted models as a column formula.
Oh, when I have a function that has many factors, I like parallel plots. Some say it is an "acquired taste." For your example, I think the multi-view (as I explained above) would be more informative for most.
I have attached an example script of the JMP 3D profiler with Mesh.
JMP has a 3D scatterplot and a Surface Profiler.
It is not too difficult with a little scripting to create a 9 plot view, or use Graph Builder: use 2 factors one for X and one for Y, making sure the one with the quadratic effect is included, then create if the other 2 factors can have discrete values like -1, 0 and 1, use one factor for X grouping and another for Y grouping then color by the modeled strength. Then instead of points use a contour.
Another possibility is to create "pictures" of the 3D plots the multiple levels of the remaining factors. and store then in a table. JMP 14 allows pictures to be used as the graphic icons, so you could create a grid of 3D views.
If your model has a secondary relationship, where the model can be reparameterized (for example maybe a combination of sealing and cooling temperature), you might be able to use it to reduce the dimensionality. However, from my experience, these are more difficult to visualize.
Thank you for elaborate answer. I think I will stay with 9 response surface graphs.
I still didn't get the answer for my first question though. I already know the 3D profiler, but I actualy want to make it 4D. I want that values of factors to determine the location of points (like 3D coordinates) and that I then colour the points, where the colour of the points represents the response (for example red for high values, blue for low). Than I would do a contour instead of points. Is there any way to colout the points in the way I described?