"Finally, it's nice if many/most of the CIs do not contain zero, hut it's not necessary. You want the CIs to be partitionable into scientifically meaningful groups. Checking that is probably best done graphically. Order the CIs in terms of the means, then plot them: --------m-------- ---m--- ------m------ etc."
did you already suspect from looking at the data that sorting the CI's for a given coefficient by increasing mean would partition the CI's in scientifically meaningful ways?
Or was that just a great hunch based on a lot of experience?
Either way, it seems to be working out quite nicely, and so far, I don't even have to do the "horizontal" dimension of the plot to get a sense of the CI width - the "vertical" sort by increasing mean is good enough to get some very clean and regular partitions.
It will take me a little bit of time to type up the a1 fold CI data so it will fit on the post page here at s.s.m. - sometime later tonight or early tomorrow ...