245
VODO

m 8 / SPLIT-PLOT DESIGN—

of
az . FACTORIAL DESIGN WITH
BLOCK-TREATMENT
CONFOUNDING
a ODNO —
ci 8.1 DESCRIPTION OF DESIGN
otic Subject heterogeneity is the rule rather than the exception in
ca behavioral research. The randomized block design described earlier
sy enables an experimenter to partially isolate the effect of subject hetero-
geneity in testing treatment effects. This is accomplished by using matched
Jade subjects or repeated measures on the same subject. In a randomized block
8) design, blocks of subjects are composed in such a way that variation among
subjects within each block is less than the variation among blocks. A
and split-plot design with repeated measures or matched subjects represents
56, an extension of this principle to experiments having two or more treatments.
| This design is appropriate for experiments that meet. in addition to the
ict. general assumptions of the analysis of variance model, the following
pren. | conditions:
rbal
Ype |. Two or more treatments, with each treatment having two or more
levels, that is, p levels of A, which is designated as a between-block or
The nonrepeated-measurements treatment, and g levels of B, which is desig-
res- nated as a within-block or repeated-measurements treatment, where
20. pandga22. . :
2. The number of combinations of treatment levels is greater than the desired
cing number of observations within each block. 7
964, 3. H repeated measurements on the subjects are obtained, each block
contains only one subject. If repeated measuremenis on the subjects are
ts in not obtained, each block contains g subjecis.
urnal 4. For the repeated-measurements case, p samples of n subjects each from
a population of subjects are randomly assigned to levels of the non-
repeated treatment (A). The seguence of administration of the repeated
ul of treatment levels in combination with one level of the nonrepeated
ype treatment is randomized independently for each block. Exception to
this procedure is made when the nature of the repeated treatment pre-
tion. cludes randomization of the presentation order.
tion. 5. For the nonrepeated-measurements case, p samples of n blocks of g

-222 subjects from a population of subjects are randomly assigned to levels
of treatment (A). After this, levels of treatment (B) are assigned randomly
to the g subjects within each block.