Augmented Designs by jizhen1947



The following problems are incorporated in this chapter:

          12a    Augmented designs
          12b    Augmented randomized complete block design
          12c    Augmented factorial designs
          12d    Spreadsheets and SAS
          12e    Canned programs

    Toward the end of any statistics class, we begin to think about the cost of doing all this
analysis! Can we save money without so much data-collecting? Can we save precious time by
using someone elses canned program?

    Many experiments provide answers that require no replication. In my studies, some tree
species or provenances (varieties) simply die in Hawaii; no need for LSDs on that! We look for
experimental designs that check out this type of lousy expectation with minimal replication and
sampling. Augmented designs (12a) provide a welcome solution to this problem.

     Augmentation is simply enlargement. An augmented design is enlarged by inclusion of
additional entries or treatments, normally without complete replication of these entries. The
replicated entries are referred to as tr, and the unreplicated as tu. The tu entries are in effect to be
compared in performance with the tr entries occurring in the same block. Adjustments of the tu
entry data are thus made to the averages of tr entries in the appropriate blocks.

     These versatile agumented designs deserve much wider use than they have had in the past. In
the field, they can result in immense savings of money and time, often with minimal loss of
information. Extensionof the augmented concept to factorial designs (2c) is increasingly common.

      SAS (SAS Institute, Box 8000, Cary, NC 27511) is one of the best canned programs
available commercially for treatment of designed experiments (12d). SAS has an extensive and
versatile set of canned programs that can address most our your queries. I stress that the
experimental biologist must first know enough statistics to lay out an intelligent experiment. Once
data are taken, you may be willing to let someone elses program analyze them for you. However,
if you are clever enough to design a very good experiment, you are probably clever enough to
carry through its analysis on spreadsheets. Then you can try out any approach you wish, graph at
will, test out random or oddball numbers anytime, and print out only what you need.

    Spreadsheets can be your entr¤e to many types of canned statistical programs (12e), and
several are discussed here with their useful websites. Data from spreadsheet can be transferred to
such programs for analysis, and this may result in savings on time, money, and your good humor.
t of doing all this
ve precious time by

tudies, some tree
on that! We look for
imal replication and

d by inclusion of
ese entries. The
ntries are in effect to be
djustments of the tu

y have had in the past. In
  minimal loss of
s increasingly common.

 ned programs
has an extensive and
  stress that the
gent experiment. Once
hem for you. However,
 ly clever enough to
oach you wish, graph at

programs (12e), and
eet can be transferred to
 and your good humor.

To top