cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Check out the JMP® Marketplace featured Capability Explorer add-in
Choose Language Hide Translation Bar
HarriBradbeer
Level II

Interpretation of leverage plots

Compare the model leverage plots below. In the first, the x-values are being offset from their actual values, which is not happening in the second.

The only difference between the two is that I've changed one number - table attached. The 214 in Factor 2 become 185 so that the two leftmost and two rightmost x-values in each plot are equal.

From reading the help guides I'm guessing the offset is something to do with the increased mean, but I don't understand why this affects the points in Factor 1, nor why the two leftmost points of Factor 2 are offset by different amounts (one +ve, one -ve) when they are equivalent. It seems to be artificially improving the model.

Any help that can be offered is much appreciated.

HarriBradbeer_1-1627916783481.png

HarriBradbeer_2-1627916960990.png

 

 

3 REPLIES 3

Re: Interpretation of leverage plots

Maybe this thread will help. It explains how the leverage plots are created. Why are the abscissa leverage values shifted in the Fit Model platform? 

 

Dan Obermiller
statman
Super User

Re: Interpretation of leverage plots

The explanations of Leverage plots Dan posted above are pretty good.  The thing you have to remember is these are plots of residuals.  So changing data points don't reflect directly on these plots.  A new model is created with the new data points and then new residuals are calculated.

"All models are wrong, some are useful" G.E.P. Box
arooqi
Level II

Re: Interpretation of leverage plots

You're creating a multiple regression model where the value of one variable affects on the prediction of the whole model. Changing the value of one factor with holding the response fixed will definitely change the CI of the factors.

Iraq Abdulwahhab