Class Notes for AB – Monday, 6^th February

Last Class

· Ex. 3.1 – contrasts are used often in practice by researchers

· Ex. 3.2 – two way analysis is correct and more informative; one-way analysis is wrong, but if used with contrasts gives 2way

· Ex. 3.3 – interaction is significant, so we cannot make a blanket statement about which usage level is lowest – it depends

· Hence, we must do a “within” or nested analysis

· Ex. 3.4 – revisiting ANOCOV from ANOVA point of view, illustrates confounding of effects, and that we must then use “LSMeans” for the diets (Type III SS)

Today’s Class

· Ex. 3.5 – SLR is more informative here than a paired t-test

· Ex. 3.6 – the importance of blocking (grouping EU into homogeneous “blocks”), one-way ANOVA is wrong, whereas RCBD analysis is more powerful – block SS comes right out of error SS, thereby increasing our power to see diff. amongst trts.

· IBD is “balanced” if conditions on p.13 are met – ex’s. 3.7, 3.8

· Ex. 3.8 – since blocks (batches) and trts. (catalysts) are confounded, use “LSMeans” not Means

· Crossover designs – very important in practice, but hard to separate out SEQ, PERIOD, TRT and CARRYOVER effects

· Ex. 3.9 – note different error term is used for SEQ, for PERIOD= TIME use Type I SS, CARRY adjusted for TRT=FORMN is signif. (p=0.0114), and so is FORMN adjusted for CARRY (p=0.0050); again use “LSMeans” for FORMN

· Ex. 3.10 another COD e.g., but with an interesting way to assess CARRY effects; note CO is NS, but one component is!