Thank you, Curt, empircism is always good.  But I was hoping this topic might have been studied systematically, that there might be a theoretical or documented benefit of one approach over the other.  Anyone?

Thank you Doug.  The comments on extensive uniform blue sky vs. marked variation in color seem well taken, I'll keep this method of choosing in mind.  My goal is to create a JPG family photo archive of the highest quality images that I can make for future use by non-technical descendants (thus it will supplement the TIF archive that holds the best quality versions of the same images but that may not be usable to novices).  As I cannot anticipate exactly how the JPGs will be used, I just want them to be the best possible, while still being of a size that can be uploaded to, say, Costco (5 MB size limit) for making enlargements. 

In general, I am often left curious as to how exactly Photoshop carries out its algorithms and how different factors influence the outcome.  So often, one read "just try different techniques and see what looks the best".  But I am always left wondering, what is the theory behind this and has it been systematically studied and worked out and published.  In so many disciplines, such as medicine, the methods of optimization has been evaluated, systematized, and fully described.  I have not yet explored what may be found in technical journals, but I'm sure much of this good stuff must be available somewhere. It would be nice to have a "How Things Work" that actually explains what Photoshop is doing under the hood.

Thanks again.

Very interesting about unlucky 7 for JPG compression, at which point no chroma subsampling is performed.  I'll keep this in mind, thank you.

Thank you Noel.  I confirm what you say, it is amazing how noise increases JPG file size.  I'll keep your advice in mind.

Thank you Michael.  As I mentioned, I will be keeping lossless TIFs of these images as well, but want the JPGs to be less than 5 MB to be most usable to unskilled posterity.  An interesting point you make that only parts of the image will be degraded by JPG compression.  So far I have converted 400 images to JPGs, and have gravitated to using Adobe JPG compression 10 and downsizing to the size necessary to be saved at < 5 MB.  This is proving to be about 4000 to 4800 pixels on the largest side.

Sounds good, I'll keep my eye out for JPG artefact at sharp edges, have not seen any at level 10 so far.

I set the goal of keeping JPGs below 5 MB for the pragmatic reason that larger files cannot be uploaded to Costco.  Admittedly, this is a lame criterion, but anyway I'm sticking to < 5 MB for JPGs.  It does add time to keep whittling down the pixel size to get the JPG to save at less than 5 MB (downsampling each time from a full size image), but hey I'm retired and these images are for the ages (or at least for my grandkids). 

As far as Curt's suggestion to strip metadata, I specifically wanted it to be present in the JPGs because I have extensively annotated the images with IPTC Descriptions, which can be very useful to have, so I am not doing a Save for Web.

Thank you all, many excellent pointers to keep in mind.