Can I use "custom test" to calculate errors for parameter combinations in a linear model
Jul 16, 2012 1:18 PM(1402 views)
I have a linear model of the form yval = grp + xval + grp*xval, where grp is categorical with two levels and xval is continuous. That I am running in JMP v. 8, with that Fit Model dialog.
JMP parameterizes it as
intercept + grp(level1) + xval + xval*grp(level1)
I want to know the estimate and standard errors for an overall intercept and slope for each of my two levels of grp. I can do this if I rerun the model separately for each level of group, but is there a way to do it with the model output from the full model? I tried using custom tests and I can get roughly the same parameter estimates as when I do two separate regressionsjmp_* , but the standard errors are very different. Which is the more valid standard error to use?