Date: Sat, 9 Mar 1996 07:41:43 -0400
Reply-To: Mitch Gallant <mgallant@PILOT.IH.NAVY.MIL>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: Mitch Gallant <mgallant@PILOT.IH.NAVY.MIL>
Subject: Pushing the ADXXVERT macro
Friends of SAS,
I am a novice SAS user who has a fair amount of experience with the regular
factorial-type experiments. I have been learing how to use SAS as a tool
for those types of designs. The area of mixture designs is however rather
new to me. I seek your help in understanding a specific macro, the
ADXXVERT macro. I have given my difficultly and lot of thought. Now I am
asking the assistance of those more experienced than I.
An innoucous example,
I wish to make the best tasting stew using water, beef, carrots, and
potatos. In this example, the water must remain at a concentration of 0.50
for all treatments. The beef, carrots, and taters can vary 0 - 0.50. I
wish to model the taste and appearance responses. Assuming I need a
quadratic model, I have tried variations in the ADXXVERT statement as in
the following code:
%ADXXVERT (stew, bee -.5 /car -.5 /pot -.5 /wat .5-.5)
PROC PRINT etc...
This line produces a data set with a number of duplicate treatments which I
think are the result of the macro treating the water as two separate of
levels of 0.50 and 0.50.
Question 1: Is there a way to format the statement to not produce these
duplicate levels? Or better stated, what statement(s) should I use to
generate a "good" design given the experiemental constraints?
Question 2: Assuming I would like the final design to include some measure
of the variation, what is the minimum design size? Do I use PROC OPTEX?
Geometrically, I visualize the tha above example as a pyramid with an
intersecting plane at water = 0.50. One approach may be to simplify my
mixture to a three component mixture, and scale my "recipes" in the
generated design by 0.50. Right now I don't fully know what implications
this scaling would have on my analysis. Also I can't find how to constrain
the macro to force three ingredients in a three mixture design to sum 0.50
instead of 1.0. Any opinions?
Thanks so much for your attention +<:)
Chemist, Naval Surface Warfare Center/Indian Head Div.
Indian Head, MD
301/743-4272 (voice mail)