You are on page 1of 15

ANALYSIS OF VARIANCE.

FACTORIAL STRUCTURE OF TREATMENTS

FACTORIAL STRUCTURE
Example: factorial structure with two factors
All the possible combinations of
each level of each factor are studied
P1

P2

P3

P4

P5

F1

F1P1

F1 P 2

F1P3

F1P4

F1P5

F2

F2P1

F2 P 2

F2P3

F2P4

F2P5

F3

F3P1

F3 P 2

F3P3

F3P4

F3P5

FACTORIAL STRUCTURE
Number of levels:
- If a linear response is expected: 2-3 levels/factor
- If a quadratic response is expected: 3-5 levels/factor

* Once decided factors and levels assign treatments at random


among Exp. Units across blocks/repetitions.
Advantages of factorial structure:
- Best way to analyze interactions.
- When interactions are not significant this is the most
efficient structure to test treatments.
Problems:
- Insufficient number of replicates in each combination of
levels of each factor?

Suplementacin conjunta con aceite de girasol


rico en oleico (AG RO) y CLA sobre la firmeza de la yema

(segn lvarez et al., 2005)

ANALYSIS OF VARIANCE OF AN
EXPERIMENT FACTORIAL SIMPLE
The survival rate of Salmonella typhimurium is
determined in response to the effects of three levels of
srbic acid and six levels of water activity. Radom block
design with 3 blocks and 18 factorial combinations of
treatments. Data analyzed are log (density St/ml) after
7 d of treatment.

3 levels SORBIC
ACID (SA)

6 LEVELS
WATER ACTIV.
(wa)

(MEAD & CURNOW)

Sorbic acid
0 ppm
Sorbic acid
100 ppm
Sorbic acid
200 ppm

aw
0.98
0.94
0.90
0.86
0.82
0.78
0.98
0.94
0.90
0.86
0.82
0.78
0.98
0.94
0.90
0.86
0.82
0.78

I
8.19
6.65
5.87
5.06
4.85
4.31
7.64
6.52
5.01
4.85
4.29
4.13
7.14
6.33
5.20
4.41
4.26
4.93
5.54

II
8.37
6.70
5.98
5.35
4.31
4.34
7.79
6.19
5.28
4.95
4.43
4.39
6.92
6.18
5.10
4.40
4.27
4.12
5.50

III
8.33
6.25
6.14
5.01
4.52
4.20
7.59
6.51
5.78
4.29
4.18
4.18
7.19
6.43
5.43
4.79
4.37
4.15
5.52

Total
8.30
6.53
6.00
5.14
4.56
4.28
7.67
6.41
5.36
4.70
4.30
4.23
7.08
6.31
5.24
4.53
4.30
4.40
5.52

ANALYSIS OF VARIANCE

(IGNORING FACTORIAL STRUCTURE OF TREATMENTS, I.E. TREATING


EXPERIMENT AS 3 BLOCKS X 18 TREATMENTS)

ss

df

ms

Blocks

0.01

0.005

Treatments

83.09

17

4.888

Error

1.77

34

0.052

Total

84.87

53

F
940***

TWO WAY INTERACTION TABLE FOR SORBIC ACID AND


WATER ACTIVITY LEVELS
Water activity

Sorbic acid

Sorbic acid

0.98

0.94

0.90

0.86

0.82

0.78

totals

8.30

6.53

6.00

5.14

4.56

4.28

5.80

100

7.67

6.41

5.36

4.70

4.30

4.23

5.44

200

7.08

6.31

5.24

4.53

4.30

4.40

5.31

aw totals

7.68

6.42

5.53

4.79

4.39

4.31

5.52

Total ss treatments = 83.09


ss main effect of water activity = 79.05
ss main effect of sorbic acid = 2.31

ss interaction = 1.73 = measures the variation between the 18


treatment combinations which is not accounted for by the two main
effect ss = 83.09 (79.05 + 2.31)

FACTORIAL STRUCTURE-ORDER SAS


proc glm;

class SA WA;
model logSALM = SA WA SA*WA;

means SA WA/t bon...;


Lsmeans SA*WA/stderr pdiff1;

run;
1In

case of unbalanced results

TWO WAY INTERACTION TABLE FOR SORBIC ACID (SA)


AND WATER ACTIVITY (WA) LEVELS
Full analysis:

ss.

d.f.

m.s.

Blocks

0.01

0.005

Water activity

79.05

15.810

304.0***

Sorbic acid

2.31

1.155

22.2***

aw x sa

1.73

10

0.173

3.33**

Error

1.77

34

0.052

Total

84.87

53

TWO WAY INTERACTION TABLE FOR SORBIC ACID (SA)


AND WATER ACTIVITY (WA) LEVELS

Sa1
Sa2
Sa3

1,2

0,8

0,8

0,6

0,6

0,4

0,4

0,2

0,2

0
1 1

22

33

4 4 5

Interaction NS

56

wa

Sa1

log (J/ml)

log (J/ml)

1,2

Sa2

Sa3

Interaction signif.

wa

EFECTO DE LA ADICIN AL PIENSO DE DISTINTAS FUENTES DE


AG -3 SOBRE EL CONTENIDO TOTAL DE EPA Y DHA EN EL
HUEVO (MG/HUEVO 60 G)

mg epa+dha huevo

300

250
200
150
100
50
0
0

2.5

7.5

10

12.5

15

17.5

20 22.5 25 27.5

g/kg w-3 pienso


DHA

EPA

soja

linaza

(Cachaldora et al., 2005 a,b, y 2006)

FACTORIAL STRUCTURE: RESULTS PRESENTATION


If AB NS: means of main effects
If AB significant: two-way table showing treatment means of all the
combinations of levels of each factor.

If any interaction results almost significant (e.g. P = 0.12%) it can


also be analyzed: treatment A at level B1 led to an increment over
a control diet a 12.9% higher than in B2; this difference had a SE of
5.87

TABLE OF TREATMENT MEANS


WATER ACTIVITY
Sorbic acid

0.98

0.94

0.90

0.86

0.82

0.78

sa mean

0 8.30

6.53

6.00

5.14

4.56

4.28

5.80

100 7.67

6.41

5.36

4.70

4.30

4.23

5.44

200 7.08

6.31

5.24

4.53

4.30

4.40

5.31

6.42

5.53

4.79

4.39

4.30

5.52

wa mean

7.68

SEM for values in body of table = (s2/3) = (0.052/3) = 0.131


SEM for water activity means = (s2/9) = 0.075
SEM for sorbic acid means = (s2/18) = 0.054
S = 0.228

OPTIMALEVALUE (MEAD & CURNOW)


E = no experimental units- no treatment combinations
E = 54 (6 x 3 x 3) 18 (6 x 3) = 36.
E optimal = 10-20
E > 20: excess experimental units.
Excp: cheap experimental units (insects, IV),
complex experiments (no comparisons,
small sized-effects).

You might also like