Split Plot Arrangement
The split plot arrangement is specifically suited for a two or more factor experiment.
This arrangement can be used with the CRD, RCBD, and LS designs discussed in this course.
Features of this design are that plots are divided into whole plots and subplots.
Example
Whole plots are wheat varieties (a0 to a3) and subplots are rates of a herbicide (b0 to b2).
a1 a2 a3 a0
b1 b2 b0 b2
b2 b0 b1 b0
b0 b1 b2 b1
With a split plot arrangement, the precision for the measurement of the effects of the whole plot
factor(s) are sacrificed to improve that of the subplot factor.
Measurement of the subplot factor and its interaction with the main-plot factor is more precise
than that obtained with an RCBD with a factorial arrangement.
Determining Which Factor to Use as the Whole and Subplot Factors
With the split plot arrangement, plot size and precision of measurement of the effects are not the
same for whole and subplot factors. Thus, assignment of a particular factor to either the whole
or subplot is extremely important. To make a choice, the following guidelines are suggested:
1. Degree of Precision: for a greater deal of precision for factor B than factor A, assign
factor B to the subplots and factor A to the whole plots.
2. Relative Size of the Main Effects: If the main effect of one factor (.e.g., factor A) is
expected to be much larger and easier to detect than that of the other factor (e.g., factor
B), factor A should be assigned to the whole plots and factor B to the subplots. This may
increase the chances of detecting differences among levels of factor B.
3. Management Practices: Cultural practices required by a factor may dictate use of large
plots. In such a case, such factors should be assigned to whole plots.
1
Randomization and Layout
The randomization procedure for split plots consists of two parts:
1. Randomly assign whole plot treatments to whole plots based on the experimental
design used.
2. Randomly assign subplot treatments to subplots. The randomization procedure
has no effect on assignment of subplot treatments to subplots.
Expected Mean Squares for the Split Plot Arrangement
The example to be given will be for an RCBD with factors A and B considered as random
effects.
Source of variation df Expected mean square
Replicate r-1 σ 2 + bσ γ2 + abσ R2
A a-1 σ 2 + bσ γ2 + rσ AB
2
+ rbσ A2
Error (a) = Rep x A (r-1)(a-1) σ 2 + bσ γ2
B b-1 σ 2 + rσ AB
2
+ raσ B2
AxB (a-1)(b-1) σ 2 + rσ AB
2
Error (b) = Rep x B(A) a(b-1)(r-1) σ2
Total rab-1
2
ANOVA of a Split Plot Arrangement
Data Set
Treatments Blocks (j)
Ai Bk 1 2 3 Yi.k
a0 b0 13.8 13.5 13.2 40.5
b1 15.5 15.0 15.2 45.7
b2 21.0 22.7 22.3 66.0
b3 18.9 18.3 19.6 56.8
Main plot total (Y0j.) 69.2 69.5 70.3 209.0 = Y0..
a1 b0 19.3 18.0 20.5 57.8
b1 22.2 24.2 25.4 71.8
b2 25.3 24.8 28.4 78.5
b3 25.9 26.7 27.6 80.2
Main plot total (Y1j.) 92.7 93.7 101.9 288.3 = Y1..
Rep total (Y.j.) 161.9 163.2 172.2 497.3 = Y...
Treatment Totals Table
a0 a1 ΣBk
b0 40.5 57.8 98.3
b1 45.7 71.8 117.5
b2 66.0 78.5 144.5
b3 56.8 80.2 137.0
ΣAi 209.0 288.3 497.3
3
Step 1. Calculate Correction Factor:
Y...2 / (r x a x b) = 497.32 / (3 x 2 x 4) = 10,304.47
Step 2. Calculate Total SS
ΣYijk2 - CF = (13.82 + 15.52 + 18.92 + . . . + 27.62) - CF = 516.2
Step 3. Calculate Replicate SS
ΣY.j.2 / axb - CF = (161.92 + 163.22 + 172.22) / (2x4) - CF = 7.87
Step 4. Calculate A SS
ΣYi..2 / rb - CF = (2092 + 288.32)/(3x4) -CF = 262.02
Step 5. Calculate Whole Plot SS
ΣYij.2 /b - CF = (69.22 + 69.52 + . . . + 101.92) / 4 - CF = 274.92
Step 6. Calculate Whole Plot Error SS = Error(a) SS
Whole Plot SS - A SS - Rep SS = 5.03
Step 7. Calculate B SS
ΣY..k2 / ra - CF = (98.32 + 117.52 + 144.52 + 137.02) / (3x2) - CF = 215.26
Step 8. Calculate AxB SS
ΣYi.k2/ r - CF - A SS - B SS = (40.52+ 45.72 + . . . + 80.22)/3 - CF - A SS - B SS
= 18.7
Step 9. Calculate Error(b) SS = Total SS - Rep SS - A SS - Error(a) SS - B SS - AxB SS
= 7.24
4
Step 10. Make ANOVA Table (Assuming A and B are fixed effects)
SOV df SS MS F
Replicate 2 7.87 3.935 6.53*
A 1 262.02 262.02 104.183**
Error(a) 2 5.03 2.515
B 3 215.26 71.753 119.993**
AxB 3 18.70 6.233 10.337**
Error(b) 12 7.24 0.603
Total 23 516.12
LSD’s for Split Plot Arrangement
1. To compare two whole plot means averaged over all subplot treatments (.e.g. a0 vs. a1)
2 Error (a ) MS 2(2.515)
tα / 2,err ( a ) df = 4.303 = 2.79
rb 3x 4
2. To compare two subplot means average over all whole plot treatments (.e.g. b0 vs. b3)
2 Error (b) MS 2(0.603)
tα / 2,err ( b ) df = 2.179 = 0.98
ra 3x 2
3. To compare two subplot treatment means for the same whole plot treatment
(.e.g. a0b0 vs. a0b3).
2 Error (b) MS 2(0.603)
tα / 2,err ( b ) df = 2.179 = 1.38
r 3
5
4. To compare two whole plot means at the same or different sub plot treatments
(.e.g. a0b0 vs. a1b0) or (.e.g. a0b0 vs. a1b3).
2[(b − 1) Error (b) MS + Error (a ) MS ]
t 'α / 2,ab
rb
Where t’ab is a weighted estimate of t that can be calculated using the following
formula:
(b − 1) Error (b) MS * tα / 2, Err ( b ) df + Error (a ) MS * tα / 2, Err ( a ) df
t 'α / 2,ab =
(b − 1) Error (b) MS + Error (a ) MS
NOTE: This formula is to calculate t, not the degrees of freedom of t as was done for other
situations.
t’ab = (4-1)(.603)(2.179) + (2.515)(4.303) = 3.414
(4-1)(.603) + 2.515
2[(4 − 1)(0.603) + 2.515]
Therefore, the LSD = 3.414 = 2.90
3x 4
Table of Means for the example
A levels
B levels a0 a1 Mean of B
b0 13.5 19.3 16.4
b1 15.2 23.9 19.6
b2 22.0 26.2 24.1
b3 18.9 26.7 22.8
Mean of A 17.4 24.0
6
Table of Means for the example LSD #4
A levels
B levels a0 a1 Mean of B
b0 13.5 19.3 16.4
b1 15.2 23.9 19.6
b2 22.0 26.2 24.1
b3 18.9 26.7 22.8
Mean of A 17.4 24.0
LSD #3
LSD #2
LSD #4
LSD #1