j 8 INTRODUCTION TO BASIC CONCEPTS IN EXPERIMENTAL DESIGN

a EO ERA

approach which is statistical control. In some experiments it may .be possible
—hrough the use of regression procedures (see Chapter 12)—0 remove the
effects of a nuisance variable statistically. This use of statistical control is
referred to as the analysis of covariance.

.

CLASSIFICATION OF INDEPENDENT
AND NUISANCE VARIABLES

AlH independent and nuisance variables in behavioral research
can be classified in one of three general categories—organismic, environ-
mental, and task variables. In the radiation example, the independent
variable of radiation can be classified as an environmental variable. The
nuisance variables listed earlier as sex, weight, prior experience, and infec-
tious diseases are examples of organismic variables. The other nuisance
variable of temperature variation among the cages is an example of an
environmental variable. This radiation experiment does not include a
task variable. A task variable could be introduced into the experiment
by reguiring the rats to perform easy, medium, and difficult visual dis-
criminations before gaining access to food. The effect of the visual dis-

, crimination on food consumption represents an additional independent
si variable that can be classified as a task variable. In the design of experi-
ments, the above classifications may help an experimenter in listing the
nuisance variables that should be controlled.

EFFICIENCY AND EXPERIMENTAL DESIGN

An experimenter engaging in research is desirous of arriving at
valid conclusions. At the same time he hopes to accomplish this goal as
efficiently as possible. Generaliy several experimental designs can be
used in testing a statistical hypothesis. However, alternative designs that are
egually valid for testing a hypothesis are rarely egually efficient. Eficiency
of alternative research procedures may be defined in different ways. For
example, efficiency may be defined in terms of time reguired to collect data,
cost of data collection, ratio of information obtained to cost, and so on.
A discussion of relative efficiency by Cochran and Cox (1957, 31) is in-
structive. A commonly used index for assessing the relative efficiency
of two experimental designs is given by the ratio of their respective experi-
mental errors. Experimental error refers to all extraneous. variation in
dependent variable scores that tends to mask the effects of the independent
variable. The main sources of experimental error are inherent variability
in the behavior of subjects and lack of uniformity in the conduct of the
experiment.

A formula that provides insight into factors related to the efficiency
of two designs is