Animals Split plot DOE - Proposal Split Plot

[#1]

In addition to foregoing discussion regarding the Animals case and after reading two excellent jmp webcasts below, please find in attachment a proposal for a split-split plot DOE set up using Custom design. Factor "Season" is very hard to change and "Species" hard to change. We are interested in the Season and Species Fox/Coyote effect and their interaction on miles run . Agree with this proposal? 

I have two addtional questions (1) Can we observe the same animals during each season or select other coyotes and foxes during each season? (2) For analyses of split split DOE I notice that REML as well as Generalized regression is used: which analysis is preferred and why?

Recommended JMP webcast split/strip plot Split-Plot and Strip-Plot Designed Experiments 

[#2]

This does not make sense to me, @frankderuyck. Why did you design this as an experiment? Surely this is an observational study.

If the aim was to better understand split plot experiments, I suggest that you look at examples of split plot experiments.

[#3]

I agree and I had a look at split plot examples & also some webinars. That's why I think one could launch this study as a controlled experiment for we assign dedicated animals = experimental units to a season, a long time frame. In each seaon we will collect many observations/unit so this must be whole plot. Eventually one could perform treatments (food type..) on both animals that can taken up as easy to change factors in the design.

[#4]

I still don't understand why you want to design this as a split plot experiment. You should always start by thinking long and hard about the actual scientific question that the study is aiming to answer. And then use that to decide on the appropriate tools and methods to design the data collection plan.

[#5]

The study aims to answer the question whether there is a difference between run miles between a fox and an coyote? And if the result is this the same for each season? I don't understand why DOE can't be applied on specific observational studies where the factors and levels are well specified?