IGN

] the
dock
esign

ay of
which
icular
ad (2)
xpeti-
ns:

i

d on an
eatment

cts may
rt to the

sacrifice
ower in

adeguate

e popula-

observed

ury to the
rludes the

to test the
r considers

> II errors?

fficiency in

employing
nment ofa

zana, smento EM IN

SN NA SANA

SECTION 1.4 OVERVIEW OF TYPES OF EXPERIMENTAL DESIGNS

(b) Can efficiency be increased more by the use of a larger sample size
or by exercising additional experimental controls during the conduct
of. the experiment?

(c) Can efficiency be increased by the measurement of ome or more
characteristics related to the dependeni variable in order to use

,, Tegression technigues?

(d) Can efficiency be increased more by the use of a complex experimental
design that reguires considerable time to plan and analyze or by using
a simple design but a large number of subjects? If subjects are plentiful
and time reguired to obtain the data is sufficient, a simple design
utilizing a large number of subjects may be more efficient than a
complex design that involves costly planning and statistical analysis.

It should be apparent that the guestion "What is the best experi-
mental design to use?" is not easily answered. Statistical as well as non-
statistical factors must be considered. The discussion has emphasized
economic factors in the selection of a design because rules can be explicitly
stated for increasing the precision and power of an experimental methodol-
ogy; but, when efficiency is considered, such rules are difficult to formulate.

ROLE OF EXPERIMENTER AND STATISTICIAN

Jt is the conviction of the author that the selection of the best
experimental design for a particular research problem can be most expedi-
tiously accomplished when the roles of experimenter and statistician are
performed by the same person. This is essentially the same position taken
by Finney (1960, 3), who states, «to write of the 'experimenter' and the
'statistician" as though they are separate persons is often convenient;
the one is concerned with undertaking a piece of research comprehensively
and accurately yet with reasonable economy of time and materials, the
other is to provide technical advice and assistance on guantitative aspects
both in planning and in interpretation |... the statistician can produce
good designs only if he understands something of the particular field of
research, and the experimenter will receive better help if he knows the
general principles of design and statistical analysis. Indeed, the two roles
can be combined when an experimenter with a little mathematical knowl-
edge is prepared to learn enough of the theory of design to be able to
design his own experiments."

CRITERIA FOR EVALUATING AN
EXPERIMENTAL DESIGN

Many diflerent sets of criteria could be given for evaluating an
experimental design. The criteria presented by Winer (1962, 47) and
Lindguist (1953, 6) are most helpful. The following guestions, except for
number 5, were selected because they touch on the major points presented
in this chapter.

21