Hi everyone,
When I fit a model in JMP using factor coding and set the model to no intercept, I noticed that the residual normal quantile plot (QQ plot) behaves differently compared to the version with an intercept.
Specifically, the residuals and the fitted “robust fit line” in the QQ plot seem to shift vertically, and even after centering the residuals, the slope of the fitted line changes slightly. In addition, I also noticed that in some cases, the larger residuals (right side of the plot) still lie close to the fitted line, while the smaller residuals (left side) deviate more strongly.
I’m wondering how JMP internally handles residuals and the “robust fit line” calculation when the model is fit without an intercept and factors are effect coded.
Could anyone explain why this shift and slope difference occur, and how the QQ plot line and confidence curves are computed under this setting?
Thanks a lot for your help!
JMP PRO 17.1.0
