PG851: GLM--ANOVA

This blog is for the use of students enrolled in either section of PG851. It is intended to be a forum for asking, discussing, clarifying, and helping.

Tuesday, February 03, 2009

Unequal Cell Sizes

After reading pages 216-220 in the textbook on unequal cell sizes, I am still having a hard time comprehending how to tackle question 2F. Is there anywhere in our notes or in the book that more clearly addresses the issues of unequal cell sizes in factorial ANOVA? Also, is nonorthogonality an issue for an interaction variable in a 2 x 2 factorial design, or is this only the case for the highest interaction term in higher order designs? Thanks!

posted by lbanderson @ 10:25 PM  

5 Comments:

  • At 6:56 AM, Blogger Mari said…

    First to address part 1 of question 2.F. (What concerns do you have?)

    The highest order interaction in a 2 X 2 design is also the only interaction term. That is, just because there is only one interaction term does not take away from the fact that it is the highest.

    If you did not find the information on p. 216-220 of your book to be sufficiently helpful, I'd encourage you to re-read the section in chapter 3 on orthogonality. However, I would also encourage you to re-read the section in chapter 5, paying careful attention to the two statistical problems outlined by Tabachnick & Fidell on this issue.

     

  • At 7:02 AM, Blogger Mari said…

    The second part of question 2.F. requires you to think about how you might evaluate the seriousness of the issue of unequal cell size in this design.

    This is a thought question, requiring you to look at the design you are running, and think about the concerns you identify in part 1 and how big of a deal they are for this particular design.

    Reasonable people can and do disagree on such matters. What the TAs and I are looking for is that you are being thoughtful about the issues.

     

  • At 7:08 AM, Blogger Mari said…

    Finally, part 3 of 2.F. asks you what you might do to address issues of unbalanced n in your study.

    Within SPSS, statistical modifications to ANOVA for designs with unbalanced ns are available--and were both presented in your notes and discussed in class before the first exam. I would also encourage you to think of other, nonSPSS ways to address unbalanced n in this answer. That is, what could you do as a researcher if you discover you have unbalanced n?

     

  • At 4:26 PM, Blogger lbanderson said…

    Thank you very much for the thorough response!

     

  • At 4:58 PM, Blogger Mari said…

    No problem...

     

Post a Comment

<< Home