Yesterday I visited the Jouy-en-Josas site of INRA. That is INRA (Institut National de la Récherche Agronomique), not INRIA, which I have visited in the past. You will see from their sign that English dirty words have reached France …
The unit used to be on the edge of the park at Versailles, so that staff and visitors could walk in the park without paying the entry fee. But as a result of a “rationalisation”, it was combined with an animal research institute on the site of the latter. The two institutes have remained separate, except for the mathematics and statistics divisions, which have been merged and put in the former pigsty (as befits the scientists who are more equal than others). Further “rationalisation” of the universities and research institutes in and around Paris will no doubt cause further changes in the near future.
Rosemary is spending a week working with Hervé Monod. On Tuesday morning she gave the second of two seminars. The first, I am afraid I can’t tell you the title or the speaker’s name, having completely failed to find it on the website; it was about modelling metabolic pathways in bacteria.
Rosemary’s talk described A-optimal block designs (those which minimise the average variance of treatment difference estimators) with very low replication, less than 2 (so that many treatments only occur once). This situation may occur with variety trials, where only small amounts of seed of new varieties are available). Comparisons between once-occurring varieties in different blocks depend on paths running through a “core” design consisting of the treatments with replication greater than 1, and it is the choice of this core which is crucial. The picture is very complicated: sometimes a control treatment in each block is best, other times it simply wastes resources. I won’t try to say more, except for two things: in this very sparse case, the Levi graph of the block design makes things much clearer; and if you use a different criterion for optimality the answer may be completely different.
After lunch I left Rosemary and Hervé to work, and went for a walk on forest paths to Versailles and back.
Peter Cameron's Blog 