Subscribe Bookmark RSS Feed

Can I use Leverage Plots for Ordinal (Nominal) Data?

AE9999

Occasional Contributor

Joined:

Nov 29, 2016

I've noticed that you can genrate leverage plots in a DOE for Ordinal (Nominal) data, but am wondering if they are meaningful without Continuous data.  Reason being is when I try to generate manually as instructed by the "Effective Leverage Plots 2" (http://predictum.libsyn.com/) , I can do first part and regress the other factors against my response without the X I am trying to generate a Leverage plot for and I see the leverage plots.  I save my residuals, but when I try to regress X against other (Nominal) factors, I do not get options for Standard Least Squares and Effective Leverage, only Nominal Logistic/

 

Question is how does JMP generate these if I can not do using the manual method and can you use leverage plots to analyze Nominal data?

1 ACCEPTED SOLUTION

Accepted Solutions
jiancao

Staff

Joined:

Jul 7, 2014

Solution

Yes, you can use leverage plots for categorical factors.  But the x-axis values are constructed differently from the ones for continuous factors. Below are the steps according to JMP Technical Support:

  • Fit a model with all factors and calculate the residuals for this full model;
  • Fit a model a model without the categorical effect of interest, and calculate the residuals for this partial model;
  • The values for the X axis are calculated this way: [residuals for the partial model - residuals for the full model] + mean of the response variable, Y.

 Please check out the Leverage Plot Details section in the JMP documentation http://www.jmp.com/support/help/13/Leverage_Plot_Details.shtml#1502500

2 REPLIES
jiancao

Staff

Joined:

Jul 7, 2014

Solution

Yes, you can use leverage plots for categorical factors.  But the x-axis values are constructed differently from the ones for continuous factors. Below are the steps according to JMP Technical Support:

  • Fit a model with all factors and calculate the residuals for this full model;
  • Fit a model a model without the categorical effect of interest, and calculate the residuals for this partial model;
  • The values for the X axis are calculated this way: [residuals for the partial model - residuals for the full model] + mean of the response variable, Y.

 Please check out the Leverage Plot Details section in the JMP documentation http://www.jmp.com/support/help/13/Leverage_Plot_Details.shtml#1502500

AE9999

Occasional Contributor

Joined:

Nov 29, 2016

Thank you for the information.  It was most helpful.