Nesting in JMP

[#1]

Hi all! I'm having trouble translating the nested structure of my data into the windows for JMP. I have an agricultural fertilizer study, where we have a control (zero fertilizer), and then 10 or so treatments (Fertilizer 1, 2, etc.). Each treatment is at two levels (low and high fertilizer, L and H). So "level" is nested within "treatment". I'm interested in the main treatment effect ("Do fertilizers differ") and whether the low and high rates differ for different fertilizers (e.g., the treatment*level effect). How do I properly code that in JMP? Many thanks in advance!

[#2]

Perhaps I don't understand your question, are you saying that the actual low and high rates are different across the different fertilizers? If the structure is indeed nested, you cannot estimate interaction effects. You write the model as follows: Y = F + R[F]

If they are the same rates, then you have a 10x2 factorial design (plus a control group). In which case you have Y = F + R + F*R.

Since F is categorical, there are no polynomial terms. 

[#3]

Thank you statman! Yes, the low and high rates are indeed the same for each of the fertilizers. So the second model would work (Y = F + R + F*R); however, the issue is the controls. The controls have one rate (zero fertilizer) which is different from the treatment rates (which have two levels that are different from zero, say 100 kg fertilizer and 150 kg fert). So in this case an interaction is not estimable if the controls are included in the analysis. So then is the solution to leave out the controls and run those in a separate analysis? Or just relativize everything by the controls (e.g., % change relative to controls) and run a 10 x 2 factorial analysis? I suppose either if I cannot nest rate within fertilizer AND estimate interactions of rate * fertilizer? Thanks again!

[#4]

There is no "right" answer. You could use 0 as the low level of the rates (or do a 3-level design for rates), perhaps %? In reality would you ever use 0?  Why nod you need to compare to 0 fertilizer? Much depends on what questions you are trying to answer and what are the motivations for the experiment: Pick a winner, or understand causal mechanisms. Whenever someone investigates factors at more than 3 levels, I have this alter ego that pops up and says "pick a winner". I would be much more interested, if in an agricultural setting, on understanding fertilizer type and amount in the face of noise. Underlying soil fertility, soil preparation, ambient conditions, variation in seed, measurement error, etc. I suggest blocking or some other strategy to assess fertilizer effects and to determine ion those effects are consistent over changing noise.

 “Unfortunately, future experiments (future trials, tomorrow’s production) will be affected by environmental conditions (temperature, materials, people) different from those that affect this experiment…It is only by knowledge of the subject matter, possibly aided by further experiments  (italics added) to cover a wider range of conditions, that one may decide, with a risk of being wrong, whether the environmental conditions of the future will be near enough the same as those of today to permit use of results in hand.”

Dr. Deming

[#5]

Hi @cmc,

Besides the excellent guidance from @statman, you can read the JMP Help section related to nested factors, and more precisely an example with nested factors that will help you define your model (without the random role of the model term): https://www.jmp.com/support/help/en/19.0/#page/jmp/example-of-a-twofactor-nested-random-effects-mode...

Hope this answer may help you,