On Nov 11, 7:25 am, djh <halitsk...@att.net> wrote: > Is this: > > http://www.alceon.com/BodyFat5.pdf > > a good pratical example of the kind of analysis you had in mind?
Yes. Your c, e, u, and L are analogous to their height, weight, fat, and age. But your situation is more complicated, for several reasons:
1. You won't be able to ignore length the way they could ignore age -- you already know that length matters. How it affects the (c,e,u) distribution is part of the question. To answer an earlier question, I don't know if you will be able to use the current length intervals (or some others), or will need to analyze at each actual length. My inclination in such situations is to not use intervals.
2. (c,e,u)|L is probably not trivariate normal. (The "|L" means "for a given value of L"; we're talking about the conditional distribution of c,e,u, with L treated as a parameter.) What that distribution is, is part of the question.
3. As I said in my Nov 10 post, your search for suitable functions needs theoretical guidance analogous to what's in the METHODS AND MODELS section of the O-F DNA paper. Without such guidance, it's hard to say where to begin, what to consider. There are no automatic algorithmic solutions.
4. *** THE BIG PROBLEM *** You need to look separately at each cell in the 6 folds x 3 sets x 2 subsets x 2 methods = 72 cells design. Some time ago -- I can't find the post -- I suggested that you analyze only one cell (or a small number of cells?). You replied that the paper would be scientifically interesting only if all the cells were analyzed. The problem, of course, is that you can't do a proper analysis of so many cells simultaneously. You need to spend time thinking about, and looking carefully at plots of, the data for each cell individually, as if it were the only one you were concerned with. That process can't be automated. You're faced with the equivalent of writing 72 different papers like Burmaster & Murray's, each of which is more complicated than theirs because of the first three reasons, plus writing a 73rd that ties them all together.
You may want to consider reducing the scope of the inquiry.