0

Running both a Split-plot ANOVA and a MANOVA on a repeated measure analysis data set and while both methods provide statically significant results the F values are really different. I would have thought they would be closer together.

Jeff
  • 167
  • What are you considering "really different?" Remember the F-distribution has a long tail. Are the probabilities really different or the F value? – RGF Oct 13 '16 at 01:44
  • They aren't "that different." Rather the MANOVA method provides (or appears to) provide more (higher) statistical evidence. – Jeff Oct 13 '16 at 01:48

1 Answers1

1

In split-plot designs you are usually assuming a correlation and in classic MANOVA the assumption is independence between groups. If you work through the various sum of squares and use the right denominators MANOVA can give larger F. I do not believe this is is necessarily the case. However, generally the assumption of independence will result in smaller sum of square errors and a larger F value.

RGF
  • 201