0
$\begingroup$

if A is the Group reference level & Noun is the Class reference level, is this summary telling us GroupC is significantly different from GroupA at only the level of nouns(intercept)? Or is it an overall sig difference considering all levels of the second categorical predictor?

enter image description here

$\endgroup$
1
$\begingroup$

The estimate of Group C tells you the predicted difference between Group A and Group C, at the reference level of Class (Noun). Likewise the estimate of Group B shows you the predicted difference between Group A and Group B, at the reference level of Class (Noun). The estimate of Class (Adjective) tells you the predicted difference between Noun and Adjective, at the reference level of Group (A).

If you want to look at the differences between Group levels under adjective Class level, you should interact Group with Class.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thank you for putting me on the right track. I am interested in the predicted differences between groups regardless of class. Do I have to dummy code Class for this? $\endgroup$ – Acer acer Jun 26 '19 at 16:11
  • $\begingroup$ You need to change the contrasts for Class. Sum or deviation coding should be fine. $\endgroup$ – user139190 Jun 26 '19 at 18:56
  • $\begingroup$ I could not know how to achieve that $\endgroup$ – Acer acer Jun 26 '19 at 19:10
  • 1
    $\begingroup$ Consider this guide for an explanation (talklab.psy.gla.ac.uk/tvw/catpred) and this guide for the implementation (mypolyuweb.hk/~sjpolit/coding_schemes.html). $\endgroup$ – user139190 Jun 26 '19 at 20:07
  • 1
    $\begingroup$ As mentioned in the second link (where they also show you the code), treatment coding reveals simple effects (what you currently have and the (default) contrasts that I described in my answer) and deviation coding reveals main effects (what I assume you desire for Class). If you deviation code Class, the intercept will no longer represent the predicted value for Group A and Class Noun, it will represent the predicted value for Group A at "mean" of Class. The intercept value should therefore be somewhat different from what you had in your original question. $\endgroup$ – user139190 Jun 27 '19 at 14:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.