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.

Wednesday, March 09, 2011

Screening Designs & ANCOVA

1.) Could you explain what it means for the interactions in screening designs to be "confounded"?

2.) In ANCOVA, could you explain what it means that with correlated covariates that regression will "blow up"?

Thank you!

posted by bfullmer @ 12:24 PM  

1 Comments:

  • At 9:54 PM, Blogger David said…

    1) It means that because your design isn't fully factorial, you're giving up your ability to test for some of the interactions. That is, you can't distinguish between some interaction and main effects (they are "confounded"). See p. 552-553 in the text for more info.

    2)Essentially, you are losing power for no real change in DV estimation. Also, it can cause computational difficulties. See section 8.3.2.2 in the book for more (p. 387).

     

Post a Comment

<< Home