Jocelyn Wen This means there is only a vanishingly small probability that these differences could have been produced by sampling error. So, we reject the idea that the differences between our means could be explained by chance. Instead, based on the evidence shown the experimental manipulation caused the difference.

Good luck.

Jocelyn Wen This means there is only a vanishingly small probability that these differences could have been produced by sampling error. So, we reject the idea that the differences between our means could be explained by chance. Instead, based on the evidence shown the experimental manipulation caused the difference.

Good luck.

You can find examples in old Design of experiments or Anova Textbooks. I haven't seen a problem like this in at least 30 years. May the force be with you. Btw you might also try this link: Statistics Problem Solver (Problem Solvers) | The Staff of REA | download (1lib.us) Best, David Booth

Hello Jocelyn,

There is an apocryphal story about the class that was made to show, by hand, the inversion of a 7 x 7 matrix. When the instructor asked, "What did you learn from this assignment?," one student (who was later celebrated in song by the rest of the class) replied, "One does not learn a d**n thing from inverting a 7 x 7 matrix by hand!"

This sounds like a bad course assignment.

The answer is, you'd need at least one additional piece of information in order to compute the SS for the two error terms (a single error term is used for the test of B and AB, as you suggest). If you had the sums of the 20 cases (B1 + B2 + B3), you could compute SS(A_resid), then, by subtraction, you could find the SS(within_resid). BTW, I think you reversed SS & MS for AB in your query.

Good luck with your work.