In the Fit Least Squares analysis, My model has a categorical variable, "Section" with three values "1", "2", and "3," and one continuous variable, "Cycles." I have three questions.
- Why doesn't the Parameter Estimates table represent the model? It gives me no information about Section "3".
- Can I change the options in JMP so the analysis output excludes the intercept coefficient and just returns Section "1", "2", and "3" coefficients as 0.655, 0.789, and 0.804?
- How do I interpret the total standard error for a given Section? Do I just stick with the Std Error in the LSM table, or do I add up the Std Error of the parameter estimates? i.e Std Error (Intercept) + Std Error (Section[#]) + St Error (Cycles)?
I'll provide a few answers.
1. JMP will leave out one value of a nominal variable in a regression model and report the coefficients for all the others - each coefficient is interpreted as the effect of that category relative to the average of all categories. This is necessary, since the model would be over-identified if all categories were used as factors. You can see the coefficient for all categories, by clicking on "Expanded estimates" after running the model. In the case where there are 2 categories, say 0 and 1, JMP will give a coefficient for category 0, and the difference in the response variable between the two categories will be 2x that coefficient (expanded estimates will show category 1 has a coefficient with the same absolute value, but the opposite sign).
2. To exclude the intercept, just check the box "no intercept" when creating the Fit Model.
3. I'm not exactly sure what you are asking here. But if you want the overall standard error for the model, use the Standard Error reported in the summary of the fit (not any individual coefficient, but the standard error of the residuals from the model).
