I am helping a colleague analyze some data, and he set up his experiment in a simple manner, but one in which I have never seen. He had reef corals in aquaria and wanted to see under which conditions they grow best. Normally, you would have many tanks, each with its own set of conditions, but he used huge tanks (to make them more "reef-like") in which conditions varied WITHIN tank. I am specifically referring to light. He basically positioned the corals at three depths so that they would receive different light levels. To me, this would appear to be a classic example of a split plot design, but what's strange to me is that there is really only the tank (whole plot) and light within tank as the subplot, whereas most examples I've seen are one treatment effectively nested within another. After watching a fantastic webinar on JMP's mixed model platform by Jian Cao (https://www.jmp.com/en_us/events/ondemand/mastering-jmp/linear-mixed-models-using-jmp-pro.html), which I have watched no less than 5-7 times in its entirety, it would appear that I would simply need to nest tank within light under the "random" mixed-model dialog to accommodate the split plot design. The design is actually more complicated because there is a temporal element, and repeated measures undertaken on corals that were assessed at different time points, but JMP can handle that as long as you use a repeated measures platform that permits random effects (e.g., unequal variances but not unstructured).
What I am confused about (assuming my nesting of tank within light was correct), is the output: JMP automatically nested my repeated measure (the coral fragments assessed over time) within my random effect (see image). And it has been presented as light(tank) rather than tank(light). Why does the repeated measure get nested within the split plot, and is the fact that the Wald p-value is highly significant telling me that my variation due to random and repeated factors is so high as to be problematic?
