support the idea

[John P. Morgan]

Hi Nam-Ky,

If we are succesful the database will be an asset to many groups, including 
members of the combinatorial, statistical design, and experimental communities.
Each design will have a lot of information attached to it; rarely will all of 
that information be of use or interest to any particular individual. The 
database makes the information available; the user selects the information 
desired. 

For users of statistical design, you are surely right that some cannot 
distinguish between "A" and "E". But you and I would likely agree that not all 
users are unable to think beyond "A".  And that there are many relevant 
statistical aspects to a design other than "A". 

The database will provide plenty of statistical information. The site will also 
provide documentation to explain that information. Those who are so inclined may 
read, think, and choose the information they wish to consider.

Those who wish to consider only "A" can simply ask for the A-best design for 
their setting. The fact that other information is available need not concern 
them.

Often it is a good idea for an experimenter to consult with a statistician about 
their design. We hope our resource will also be useful to statisticians in this 
role. The statistician will be able to employ various aspects of the statistical 
information available, as appropriate, to select a good design for the 
experimenter. 

For the use of this resource in selecting a design for an experiment, and in the 
context of the questions you have raised over several postings, here's a summary 
of my current view of what we are doing. In selecting a design for an 
experiment, each user will decide which criteria to employ. If it's just "A", 
that's okay. If more than "A", that's okay, too. We will provide information. 
Information provides potential, but does not dictate practice.

Another statistical role we envision for our resource is in advancing the study 
of optimal experimental design. Centralizing information and making it available 
can often be a stepping stone to advancing knowledge. In this regard it may help 
in bringing insight to the problems that you consider more important. But only 
time will tell. 

You are right that row-column designs are also an experimentally important 
class. In the future we hope to include them as well. It isn't that big a step 
from where we are, but this first step with simple block designs has been a 
pretty big one. Once we get this right, other steps will follow. 

Cheers, JP

 
>Date: Tue, 9 Sep 2003 23:07:15 -0700 (PDT)
>From: Nam-Ky Nguyen <nkn at designcomputing.net>
>
>It seems you me the type of database you mentioned
>will be more useful to combinatorial mathematicians
>than to experimenters. Experimenters can't distinguish
>between the A-efficiency and E-efficiency of an IBD.
>However, they can distinguish an IBD and a row-column
>design. The latter is more appropriate their
>experiments as in reality heterogeneiy are in two
>directions not one. 
>
>There are other things I consider more important than
>the database such as the tighter bound for
>(alpha-resolvable) IBDs which are not BIBD and bound
>for RCDs and I am quite sure that here is the area
>combinatorial mathematicians can contribute.
>