On Jul 24, 4:31 pm, djh <halitsk...@att.net> wrote: > Thanks very much for taking the time to consider this case. > > If this were a case in isolation, with no meaning as part of a larger > paradigm, then I would be willing to dismiss it immediately as > "significant but trivially significant", as per your judgment. > > But once the drafting of Paper I gets under way, you will see why this > case IS part of a larger and very interesting paradigm, so I hope you > will be willing to reserve final judgment on it until you see its role > in the "larger picture". > > More general question - I assume the test you just provided can be > applied any time I have questionable cases like this one, i.e. cases > of "virtual non-overlap". > > Is that correct? > > Thanks again, Ray.
That's correct. Howevber, I should also point out that the qualifier "For the approach you're taking" in my post was not vacuous. There are other ways to compare such pairs of curves. Have you lookied at all the plots? Is linear regression of the SEs on the length indices appropriate in all cases?