topic Re: Problem with Parameter Estimates in a Regression Interaction in Discussions

Problem with Parameter Estimates in a Regression Interaction

jstrut — Sun, 28 May 2017 06:15:57 GMT

Hello! I am having difficulty looking into what drives an interaction I have found. The interaction is between Age and Group. Since Age is a continuous variable, I have used a regression to look into it. I would like to do individual t-tests within the regression to see what is driving the interaction. I believe this would be under the "Parameter Estimates" section of the report.

However, for whatever reason, JMP does not include all three groups in the Parameter Estimates. It always seems to leave out a group. I tried the model with groups A, B, and C, and it gave me estimates for A and B. When I tried it with just groups B and C, it only gave me estimates for B. And when I tried groups A and C, it again only gave me estimates for A. Why is this? Can anybody help me? Thank you!

Re: Problem with Parameter Estimates in a Regression Interaction

dale_lehman — Sun, 28 May 2017 13:28:25 GMT

The mathematics requires that one group be omitted since an observation that does not belong to any of the other groups must belong to that particular group (rendering the matrix noninvertable). In order to see all the regresssion coefficients, click on the red triangle at the top of the model output and ask for the Expanded Estimates and it will show the coefficients for all the groups (the missing group coefficient is always equal to 0 - sum of the other coefficients).

Re: Problem with Parameter Estimates in a Regression Interaction

jstrut — Sun, 28 May 2017 18:35:16 GMT

That was exactly what I was looking for. Thank you so much for the help and the explanation!