How to draw heatmap to visualize main effect and quadratic effect found in definitive screening design

[#1]

Hi JMP community,

I am using DSD to find the main effect as well as quadratic effect among my factors. Specifically, my responses are multiple numerical parameters, and my factor inputs include both numerical and categorial values. Now, I have run the experiments and feed into the model. Because I am interested in the effects of all these factors to my multiple responses simultaneously, I am using prediction profiler plot to visualize the modeling results by assigning different importance levels to each response. (see attachment 1). And the model successfully predicts the best factor combinations that would generate the ideal response profiles based on my defined importance level. 

But besides this prediction profiler, is there other way to draw a heap map to show regression coefficients of main and quadratic factor effects to each response? (such as the image I attached here). So that I can investigate which effect is more important to my response.

[#2]

You can use Graph Builder to produce the heat map.  Here is a simple example, using the Big Class data table.  But it shows 2 categorical columns being used as the X and Y axes and a numeric as the Color column, which give you the desired heat map.

[#3]

Hi @Elsie24,

Welcome in the Community !

1. One possible option without manual work is to enhance your Profiler by coloring your factors in a red scale depending on their importances for each response.
   In order to do this, click on the red triangle next to "Prediction Profiler", click on "Assess Variable Importance" and then choose a relevant option for you. For unconstrained designed experiments (like in the case of a DSD), "Independent Uniform Inputs" is a relevant choice. A new panel "Variable Importance: Independent Uniform Inputs" is now present, and if you click on the red triangle next to it, you'll find the option "Colorize Profiler". When clicking on it, your factors on the Profiler will be in shades of red (the more red, the higher the importance of the factor on the response) :
   
2. A second option to match the snapshot provided would require manual work. 
   First, you need to create a table with all coefficients from active terms in the model for each response ; you can do this by right-clicking on "Parameters Estimates" and then click on "Make Combined DataTable". A new datatable is created with all necessary information to use in GraphBuilder. 
   If you use the column "Y" as Y in Graphbuilder, the column "Term" as X and "Estimate" as Color property, and then using a Heatmap graph style, you end up with something similar as you expected : 
   

   I used the Local Data Filter to remove the intercept estimate from the Heatmap, as its values are very high compared to other coefficients, so it may bias the colors used and make all terms similar compared to the intercept terms.

I hope this answer will help you,

[#4]

@Victor_G Thanks for your detailed explanation! I'm working on your second method. However, to only investigate "main effect" as well as "quadratic effect", instead I have to use "Fit Model", "Standard Least Squares", and then manually construct model effects. The generated models however, only list statistical parameters, like t ratio, P-value, etc (see attachment), but without regression coefficients. How should I get this parameter as well so I can draw exactly same heat map as I attached peviously? Many thanks~ 

[#5]

Hi @Elsie24,
You do have access to regression coefficients, these are the values listed in the "estimates" column from your snapshot. If you have categorical effects, you may also consider using the "Expanded estimates" so that you have the regression coefficients for all levels of your categorical factors (red triangle next to your response model fit, then "Estimates" and click on "Expanded estimates").

[#6]

Thanks again! I really appreciate your answer!