How to extend the plot of predicted mean values for interaction effects in Proportional Odds Ordinal Logistic Regression?

1

Dear All,
My question to you is quite short and I am only asking for help in using the ‘rms’ package to draw the third plot of the response variable predictions for the LOCATION*MATRIX interaction effect (i.e. I need an additional plot with two variable factors as I may show you in the bottom of this letter) in the example given by Dr. Charles Holbert:

based on a brilliant idea of the Proportional Odds Ordinal Logistic Regression.
I confess that I’ve spent many hours alone and I can’t complete this drawing. Then if you think that my question makes sense, then I kindly ask you to help me to write an appropriate R code solution. Of course, in a planned paper situation, your input will be quoted.
I think this additional plot would complement a lot in terms of the graphical presentation of these predictions and I will be so obliged for your help.
With words of respect, Andrzej Tukiendorf.

2

An interaction effect per se, when there is a second order interaction, is an estimate of a double difference. You can do that with the rms package contrast.rms function whose help file has examples. But I suspect you mean to get a more typical plot that doesn’t show differences but shows predicted values (e.g., means) over combinations of levels of two interacting factors. There are examples of this in the help file for the rms Predict function.

3

Yes, exactly I need this, what you write, Professor. I will look at these examples. Thank you so much.