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; } } }

parameterisation options in ANOVA

gemmahodgson

So how do you set the parameterisation used in the ANOVA model for categorical variables - i.e. say you have a variable that has 2 levels, JMP will report the parameter estaimte for one of them and the other is assumed to have coefficient of zero. Other software I use allows you to choose which one is set to zero, i.e. first or last. Or instead set the sum of the coefficients to zero so you get a coefficient for each level of the variable. Can you do this in JMP - I couldn't see where this option is so I wondered if it was a default setup that the last level of the coefficient is chosen as the reference?

Accepted answers

MRB3855

Hi @gemmahodgson : The default parameterisation in Fit Model is the "set the sum of the coefficients to zero so you get a coefficient for each level of the variable" as you describe. But, it doesn't show one of them (it is the negative sum of the others). You can see them all if you choose to "expanded estimates" in the pull down menu of opitons (red triange at top).

What you describe as the "last level of the coefficient is chosen as the reference" is the "indicator parameterisation estimates" option. And for up through v18 there is no option to choose the reference level. I can't speak for v19.

All comments

MRB3855

gemmahodgson

Aha! Very helpful thank you

rcast15

@gemmahodgson I believe if you go to Value Order in Column Properties you can set whatever level you want as your last level and Fit Model will then use this as your reference group.