Psych 610 Handout #22, p. 1
Prof. Moore
SPECIAL TESTS FOR THREE (AND HIGHERWAY) DESIGNS
I. All two-way tests and one-way tests discussed previously can be applied to collapsed (pooled)
data from higher way designs.
* Can run single df contrasts on marginal means
* Can run single df contrasts on individual cell means
* Can run two-way single df contrasts on pooled two-way matrices (i.e., CB, CP, BP)
* Can run simple main effects on pooled two-way matrices
* Can run partial interactions on pooled two-way matrices
II. Can extend notions of two-way and partial interaction contrasts to 3 (and higher) way designs if
it makes sense.
III. New test: simple interaction effect
Asks whether there is an interaction between two factors, say CP, at a particular level of a third
factor, say B.
Let’s compute SSCP@RIR =
541250
40
000,650,21
40
41002200
@][
22
==
+
=RIRC
589500
20
000,790,11
20
150020002300500
@][
2222
==
+++
=RIRP
000,637
10
000,370,6
10
10001300....900100
@][
2222
==
+++
=RIRCP
125,496
80
6300
@][
2
==RIRT
SSCP@RIR = [CP] – [C] – [P] + [T] = 2375
MSCP@RIR = 2375/3 = 791.67
52.1
83.520
67.791 ==F
ns.
Conclude: For RIR’s, there is no interaction between period and
cue. A-V is better than A at all levels by about the same
amounts.
