PlSc 724 - FACTORIAL EXPERIMENTS
Factor - refers to a kind of treatment.
Factors will be referred to with capital letters.
Level - refers to several treatments within any factor.
Levels will be referred to with lower case letters.
A combination of lower case letters and subscript numbers will be used to designate individual treatments (a0, a1, bo, b1, a0b0, a0b1, etc.)
Experiments and examples discussed so far in this class have been one factor experiments.
For one factor experiments, results obtained are applicable only to the particular level in which the other factor(s) was maintained.
Example: Five seeding rates and one cultivar.
A factorial is not a design but an arrangement.
A factorial is a study with two or more factors in combination.
Each level of a factor must appear in combination with all levels of the other factors.
Factorial arrangements allow us to study the interaction between two or more factors.
Interaction – 1) the failure for the response of treatments of a factor to be the same for each
level of another factor.
2) When the simple effects of a factor differ by more than can be attributed to
chance, the differential response is called an interaction.
Examples of Interactions
012345a0a1Factor ADependent variableb0b1012345a0a1Factor ADependent variableb0b1
No interaction (similar response) Interaction (diverging response)
1
012345a0a1Factor ADependent variableb0b101234567a0a1Factor ADependent variableb0b1
Interaction (crossover response) Interaction (converging response)
Simple Effects, Main Effects, and Interactions
Simple effects, main effects, and interactions will be explained using the following data set:
Table 1. Effect of two N rates of fertilizer on grain yield (Mg/ha) of two barley cultivars.
Fertilizer Rate (B)
Cultivar (A)
0 kg N/ha (b0)
60 kg N/ha (b1)
Larker (a0)
1.0 (a0b0)
3.0 (a0b1)
Morex (a1)
2.0 (a1b0)
4.0 (a1b1)
The simple effect of a factor is the difference between its two levels at a given level of the other factor.
Simple effect of A at b0 = a1b0 - a0b0
= 2 - 1
= 1
Simple effect of A at b1 = a1b1 - a0b1
= 4 - 3
= 1
Simple effect of B at a0 = a0b1 - a0b0
= 3 - 1
= 2
Simple effect of B at a1 = a1b1 - a1b0
= 4 - 2
= 2
2
012345a0a1Factor AWeightb0b1012345b0b1Factor BWeighta0a1
The main effect of a factor is the average of the simple effects of that factor over all levels of the other factor.
Main effect of A = (simple effect of A at b0 + simple effect of A at b1)
2
= (1 + 1)/2
= 1
Main effect of B = (simple effect of B at a0 + simple effect of B at a1)
2
= (2 + 2)/2
= 2
The interaction is a function of the difference between the simple effects of A at the two levels of B divided by two, or vice-versa.
(This works only for 2 x 2 factorials)
A x B = 1/2(Simple effect of A at b1 - Simple effect of A at b0)
= 1/2(1 - 1)
= 0
or
A x B = 1/2(Simple effect of B at a1 - Simple effect of B at a0)
= 1/2(2 - 2)
= 0
3
Example with an interaction:
Table 2. Effect of two N rates of fertilizer on grain yield (Mg/ha) of two barley cultivars.
Fertilizer Rate (B)
Cultivar (A)
0 kg N/ha (b0)
60 kg N/ha (b1)
Larker (a0)
1.0 (a0b0)
1.0 (a0b1)
Morex (a1)
2.0 (a1b0)
4.0 (a1b1)
The simple effect of a factor is the difference between its two levels at a given level of the other factor.
Simple effect of A at b0 = a1b0 - a0b0
= 2 - 1
= 1
Simple effect of A at b1 = a1b1 - a0b1
= 4 - 1
= 3
Simple effect of B at a0 = a0b1 - a0b0
= 1 - 1
= 0
Simple effect of B at a1 = a1b1 - a1b0
= 4 - 2
= 2
012345b0b1Factor BWeighta0a1 012345a0a1Factor AWeightb0b1
4
The main effect of a factor is the average of the simple effects of that factor over all levels of the other factor.
Main effect of A = (simple effect of A at b0 + simple effect of A at b1)
2
= (1 + 3)/2
= 2
Main effect of B = (simple effect of B at a0 + simple effect of B at a1)
2
= (0 + 2)/2
= 1
The interaction is a function of the difference between the simple effects of A at the two levels of B divided by two, or vice-versa.
(This works only for 2 x 2 factorials)
A x B = 1/2(Simple effect of A at b1 - Simple effect of A at b0)
= 1/2(3 - 1)
= 1
or
A x B = 1/2(Simple effect of B at a1 - Simple effect of B at a0)
= 1/2(2 - 0)
= 1
Facts to Remember about Interactions
1. An interaction between two factors can be measured only if the two factors are tested together in the same experiment.
2. When an interaction is absent, the simple effect of a factor is the same for all levels of the other factors and equals the main effect.
3. When interactions are present, the simple effect of a factor changes as the level of the other factor changes. Therefore, the main effect is different from the simple effects.
5
Example of ANOVA for a 2x2 Factorial
Table 1. Data for the RCBD analysis of a 2 x 2 factorial arrangement.
Treatments
Replicate
a0b0
a0b1
a1b0
a1b1
Y.j
1
12
19
29
32
92
2
15
22
27
35
99
3
14
23
33
38
108
4
13
21
30
37
101
Yi.
54
85
119
142
400=Y..
Step 1. Calculate Correction Factor
000,102*2*440022..===rabYCF
Step 2. Calculate Total SS
0.170,1)37...141512(SS Total22222=−++++=−=ΣCFCFYij 6
Step 3. Calculate Replicate SS
5.322*2)1011089992(Rep22222.=−+++=−=ΣCFCFabYSSj
Step 4. Partition Treatment SS
Step 4.1. Make Table of Treatment Totals
Table . Table of treatment totals.
a0
a1
3B
b0
54
119
173
b1
85
142
227
3A
139
261
400
Step 4.2. Calculate A SS
25.9302*4)261139(ASSA 222=−+=−=ΣCFCFrb
7
Step 4.3. Calculate B SS
25.1822*4)227173(BSS B222=−+=−=ΣCFCFra
Step 4.4. Calculate A x B SS
4.0SS BSSA CF4)14211985(54SS BSSA CFabSS AxB22222=−−−+++=−−−=Σr
NOTE: A SS + B SS + A x B SS = Treatment SS
Step 5. Calculate Error SS
Error SS = Total SS - Rep SS - A SS - B SS - A x B SS
= 21.0
Step 6. Do the ANOVA
SOV
df
SS
MS
F (assuming A and B fixed)
Rep
r - 1 = 3
32.5
10.833
Rep MS/Error MS = 4.64*
A
a - 1 = 1
930.25
930.25
A MS/Error MS = 398.679**
B
b - 1 = 1
182.25
182.25
B MS/Error MS = 78.107**
A x B
(a - 1)(b - 1) = 1
4.00
4.00
AxB MS/Error MS = 1.714
Error
(r - 1)(ab - 1) = 9
21.00
2.333
Total
rab - 1 =15
1170.00
8
Step 7. Calculate LDS’s (0.05)
Step 7.1 Calculate LSDA
7.12*4)333.2(2262.2rbMSError 2tA LSD.05/2;9df===
Mean of treatment A averaged across all levels of B.
Treatment
Mean
a0
17.4 a
a1
32.6 b
Step 7.2 Calculate LSDB
7.12*4)333.2(2262.2raMSError 2tB LSD.05/2;9df===
Mean of treatment B averaged across all levels of A.
Treatment
Mean
b0
21.6 a
b1
28.4 b
9
Step 7.3 Calculate LSDA x B
4.24)333.2(2262.2rMSError 2tAxB LSD.05/2;9df===
Mean of the interaction of A and B. Mean of the interaction of A and B.
Treatment
Mean
a0b0
13.5 a
a0b1
21.3 a
a1b0
29.8 a
a1b1
35.5 a
Factor B
Factor A
b0
b1
a0
13.5 a
21.3 a
a1
29.8 a
35.5 a
010203040a0a1Factor AYieldb0b1
You can see from the figure above that the two lines are nearly parallel. This indicates that B is responding similarly at all levels of A; thus, there is no interaction.
10
Example of a CRD with a 4x3 Factorial Arrangement
Given there are 3 replicates, the SOV and df would be as follows:
SOV
Df
A
a-1 = 3
B
b–1 = 2
AxB
(a-1)(b-1) = 6
Error
By subtraction = 24
Total
rab-1 = 35
Example of a Latin Square with a 3x2 Factorial Arrangement
What would be the size of the Latin Square?
Answer: 6
The six treatments would be all combinations of A and B (a0b0, a1b0, a2b0, a0b1, a1b1, a2b1).
The ANOVA table would be as follows:
SOV
Df
Row
ab-1 = 5
Column
ab-1 = 5
A
a-1 = 2
B
b-1 =1
AxB
(a-1)(b-1) = 2
Error
(ab-1)(ab-2) = 20
Total
(ab)2 –1=35
Note that r=ab
Example of a RCBD with a 4x3x2 Arrangement
Given there are 5 replicates, the ANOVA would look as follows:
SOV
Df
Rep
r-1 = 4
A
a-1 = 3
B
b-1 = 2
C
c-1 = 1
AxB
(a-1)(b-1) = 6
AxC
(a-1)(c-1) = 3
BxC
(b-1)(c-1) = 2
AxBxC
(a-1)(b-1)(c-1) = 6
Error
(r-1)(abc-1) = 92
Total
rabc-1 = 119 11
In order to calculate the Sums of Squares for A, B, C, AxB, Ax C, BxC, and AxBxC, you will need to make several tables of treatment totals.
The general outline of these tables is as follows:
Table 1. Totals used to calculate A SS, B SS, and AxB SS.
a0
a1
a2
a3
ΣB
b0
a0b0
a1b0
b1
b2
ΣA
Remember AxB SS = ()rcabΣ2- CF – A SS – B SS
Table 2. Totals used to calculate A SS, C SS, and AxC SS.
a0
a1
a2
a3
ΣC
c0
a0c0
a1c0
c1
ΣA
Remember AxC SS = ()rbacΣ2- CF – A SS – C SS
Table 3. Totals used to calculate B SS, C SS, and BxC SS.
b0
b1
b2
ΣC
c0
b0c0
b1c0
c1
ΣB
Remember BxC SS = ()rabc2Σ- CF – B SS – C SS
Table 4. Values used to calculate Total SS, Rep SS, and AxBxC SS.
Rep 1
Rep 2
Rep 3
ΣABC
a0b0c0
a0b0c1
a0b1c0
…
a3b1c1
ΣRep
Remember AxBxC SS = ()rabcΣ2- CF – A SS - B SS – C SS – AxB SS – AxC SS – BxC SS 12
Linear Model
Yijk = : + Where: : = Experiment mean
"j = Effect of the jth level of factor A
$k = Effect of the kth level of factor B
("$)jk = A x B interaction effect
,ijk = Random error
Advantages of Factorial Arrangements
1. Provides estimates of interactions.
2. Possible increase in precision due to so-called “hidden replication.”
3. Experimental rates can be applied over a wider range of conditions.
Disadvantages of Factorial Arrangements
1. Some treatment combinations may be of little interest.
2. Experimental error may become large with a large number of treatments.
3. interpretation may be difficult (especially for 3-way or more interactions).
Randomizing Factorial Arrangements
1. Assign numbers to treatment combinations.
2. Randomize treatments according to design.
Example - RCBD with a 2x4 Factorial Arrangement
Treatment
Treatment number
Treatment
Treatment number
a0b0
1
a1b0
5
a0b1
2
a1b1
6
a0b2
3
a1b2
7
a0b3
4
a1b3
8
Rep 1
3
a0b2
7
a1b2
2
a0b1
6
a1b1
4
a0b3
5
a1b0
1
a0b0
8
a1b3
13
Interpreting Results of ANOVA Involving Interaction Terms
Interpretation should always begin with the higher level interaction terms (e.g. three-way interactions before two-interactions, etc.).
Interpretation of the main effects should never be done before interpreting the interaction terms.
The F-test for interaction terms can be significant because of two reasons.
1. True interaction.
2. Differences in magn
การแปล กรุณารอสักครู่..
