Create New Entity From PCA Analysis

Cankerz

Hey everyone, 

 

Hey everyone, First post here. I am still very new to JMP and some of the analysis methods. With that in mind here is my question:

 

I am performing PCA analysis on some products to see how varied our products are from one another and I am seeing that there is a quadrant of the PCA graph that has no points in it. I want to create a new product that would fall in that quadrant that I could then produce and test. Is there a way to tell JMP to create an entity that would be at a specific point in the Component 1 vs Component 2 graph based on the principal component equations being used?

Mark_Bailey

A useful guide for you is the Loadings Plot. It shows how the original variables are correlated with the components. It defaults to the first two components, which you are interested in.

 

Identify the quadrant that is vacant of scores. For example, let's say it is when score 1 is negative and score 2 is negative (lower left quadrant). Now imagine a line through the origin pointing towards this quadrant in one direction and away in the other direction. (In this example, it would be the identity line where score 1 = score 2.)

 

Next identify the variables in the Loading Plot that align with this line. If they point towards it, then make them strongly positive values. If the point away from it, then make them strongly negative. If they are normal to this line, then they don't matter. Verify that you currently have no such observations.

 

Hope that this suggestion helps.