302 SPLIT-PLOT DESIGN

V V NINO

TABLE 8.14-1 (continued)

SS, —- [C) - [X] < 24.500
SS,c < [AC] — [A] - [C] t [X] < 10125
Š, svoj varov Z [ACS] — [AC] — [AS] t [A] < 2375
SS,c < [BC] - [B] - [C] £ [X] < 8000
SS,sc < [ABC] — [AB) — [AC] - [BC] 4 [A] t [B] 4 [C] - [X] < 3125
SSgc x subj w.grovpe [ABCS] - [ABC] — [ABS] — [ACS] t [AB) [AC] £ [AS]
— [A] < 1875

a v zen m mm —<——<€O—

TABLE 8.14-2 Analysis of Variance Table

V a o v z uv m z »— — —<——<—<€<—

Source : SS df
1 Between subjects 12.500 np-ls7
2 A 3.125 p-lel
3 Subj w.groups 9.375 pa —1l)<6
4 Within subjects s 223.000 nplar — l) < 24
5  B 162.000 g-l:l
6  AB 6.125 (p- lXa - 1 <1
7 B x subj w.groups 4.875 pin — lXa - 1) <6
8 Cc 24.500 r-ls:l
9 AC 10.125 (p- IM -1)<l
10 C x subj w.groups 2.375 pin — iXr — 1) <6
BI BC 8.000 H (a - IA -bsl
— 12. ABC 3.125 (p- Ila - MW - 1) <1
13 BC x subj w.groups 1.875 pin — lXg — IX — 1) <6
14 Total , 235.500 npar — | < 31
a o a a ER II O Z
tp <.05.
sp <.OI.

TESTS FOR HOMOGENEITY OF
ERROR TERMS

Four sets of error terms in a type SPF-2.22 design can be tested
for homogeneity. The variances estimated by MS,,,; w.groupar dt P levels
of A should be homogeneous. Similarly, the variances estimated by
MS, x subj w.group ar At P levels of A should be homogeneous, and the same is
true for MS«,,,sj wgroup ar 804 MŠpcxsubj w.group ar: Computational proce-
dures for computing the reguired mean sguares at level a, appear in Table
8.14-3. The formulas at level a, follow the same pattern as those at level a,.
An F,,,, fatio for these partitioned error terms has the form