On Dec 7, 9:27 am, djh <halitsk...@att.net> wrote: > Here are the average SE and p values for u*e in the regression c on > (u,e,u*e), for each value of subset x Moss | (Fold,Set,Len) = > (all,all,all) > > -------------------- > subset > x u*e > Method SE p > > SxN 0.491 0.454 > SxC 0.602 0.462 > SxR 0.547 0.463 > SxC 0.512 0.478 > --------------------
I'm confused by the "subset x Method" labels in this and your two previous posts. If the Subsets are S and C, and the Methods are N and R, then the labels should be some permutation of (S,N), (S,R), (C,N), and (C,R).
Also, as I've mentioned before, the "x" is usually used to refer to factors -- e.g., Subset x Method -- not particular combinations of levels of the factors. The "x" denotes the Cartesian product of the two factors; each combination is an ordered pair, conceptually the same as ordinary (x,y) coordinates. (The parentheses can be dropped, as befits the context.)
> > Unlike the previous case of u^2 in c on (e,u,u*e,u^2), SxN DOES have > the lowest SE and the lowest p. > > Now, I'm sure you'll say the differences are meaningless. > > But there has to be some reason why the average slope Aube of the e > coefficient in c = (u,e,u*e) is giving us the striking "het" result at > method = N and sets 1, 3: > > a1 a3 b1 b47 c1 c2 > C S C S c S c S C S C S "Het" > > N1 Aube H H L L H H H H L L L L 0 > N3 Aube L L H H L L H H H H L L 0