nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Question regarding modeling in JMP 10

pamplemousse

Hello,

I am dealing with 3 continuous independent variables (e.g. D,E,F) and 1 continuous dependent variable (e.g. X).

Some time (~1 year) ago, I was able to use JMP to generate a mathematical model of the form "X = c*D*E*F", where c is just a dimensionless constant. Presently, however, I am having trouble replicating that process in JMP 10; the 'closest' I have gotten so far by playing around with the "Fit Model" settings is "X = (D-d)(E-e)(F-f)", which is still quite different.

If anyone could help by directing me towards the correct method for generating a model of the form "X = c*D*E*F", I would greatly appreciate it.

Thanks!

Accepted answers

louv

Not sure why you would want to do this but in order to obtain what you are requesting once could use the Fit Model platform and specify X as your response and the 3 way interaction D*E*F as the modeling term. Then check the no intercept button and under the red triangle deselect center polynomials and you will obtain what you are looking for.

All comments

louv

pgstats

One way to do your regression is as follows: Create a new column in your table with formula c*D*E*F where c is a new parameter (created by adding c to the list of formula parameters).with some initial value. Then use the nonlinear platform to do the regression using X as the dependent variable and the new column as the X formula.

pamplemousse

Thanks a lot! Forgetting to deselect "center polynomials" was my sole impediment.

While waiting for a reply, I constructed a new column for the product "(D*E*F)" and performed a linear regression of the form X = c(D*E*F), yielding the same regression coefficient ("c") as my preferred method.