Multifactor Experiments

• Factorial Experiments
• Split Plot Design
• Nested Design

1
 Multifactor experiments
(Factorial Experiments)
Factor:- A factor is a kind of treatment e.g
• If diet is a factor, several diets may be used,
• If baking temperature is a factor, several
temperatures will be used for baking.
• If Pesticide is a factor, several concentrations
may be testes on insects.
Level:-The term level refers to the several
treatments within any factor. In above example,
different baking temperatures are levels of the
factor temperature.
2
Notations in factorial experiments:-
• Factors are denoted by capital letters e.g
A,B,C…
• Levels of the factors are denoted by small
letters a, b, c…
• Subscript number (a1,b1,c1…) to show a level
number.
• Small letters to indicate the number of levels of
factor.
For example if a factor spray to control insects has
3 levels then we may write these levels as s1,s2,s3
and s=3 to indicate the number of levels of the
factor spray (S). 3
Treatments in Factorial experiments
In factorial experiments, the experimental
treatments are formed by combining the levels of
different factors e.g if an experiment involves
spray at three different concentrations (s=3) to
control insects and these are applied by using
two methods (m=2) then the s x m=6
treatments are
m1s1, m1s2, m1s3, m2s1, m2s2, m2s3

m1 m2

s1 s2 s3 s1 s2 s3
Advantage of using Factorial experiments
instead of series of single factor experiments

• Saving of experimental recourses

• Study of interaction among different factors

Factorial experiments under
• CRD

• RCBD

• Latin Square
{2 Factor Factorial in CRD}The yield of a chemical process is being studied. The
two important factors are thought to be the pressure (P) and temperature (T). Three
levels of each factor are selected, and a factorial experiment with two replicates
under CRD is performed. The yield of the experiment is given below

Pressure

200 215 230 Response Variable= chemical yield

90.4 90.7 90.2 Levels of factor Pressure (p)=3
Low 90.2 90.6 90.4 Levels of factor Temperature (t)= 3
# of treatments (pxt)= 9
90.1 90.5 89.9 # replications (r)= 2
Temperature

Medium 90.3 90.6 90.1 # observations (pxtxr)= 18

90.5 90.8 90.4

High 90.7 90.9 90.1
 Pressure (S.O.V) PxT table of Totals
200 215 230 Pressure(P) Pressure
90.4 90.7 90.2
Low Temperature (T) 200 215 230 TOTAL
90.2 90.6 90.4

Temperature
Low
Temperature

PxT 180.6 181.3 180.6 542.5

90.1 90.5 89.9
Medium Medium 180.4 181.1 180.0 541.5
90.3 90.6 90.1 Error
High 181.2 181.7 180.5 543.4
90.5 90.8 90.4 TOTAL
High TOTAL
90.7 90.9 90.1 542.2 544.1 541.1 1627.4

Correction Factor=(G.T)2/p×t×r= (1627.40)2/18 = 147135.04

Total SS =(90.4)2+(90.2)2 …(90.1)2 – CF= 1.298
(542.2) 2 (544.1)2 (541.1)2
SS = + + − CF = 0.768
P 6 6 6
(542.5) 2 (541.5) 2 (543.4) 2
SST = + + − CF = 0.301
6 6 6
2 2
(180.6) (180.4) (180.5) 2
SS subtable P  T = + + − CF = 1.138
2 2 2
SSPxT = SSsubtable PxT- SSP – SST=0.069

SSError= 0.160 (By Subtraction) 7

 Source Of Variation Degree of Sum of Mean Sum Fcal Ftab
(S.O.V) Freedom Squares of Squares
(DF) (SS) (MSS=SS/df)

Pressure(P) 3-1 =2 0.768 0.384 21.59* F.05(2,9)=4.26

Temperature (T) 3-1 =2 0.301 0.151 8.47* F.05(2,9)=4.26
P xT 2x2 =4 0.069 0.017 0.97ns F.05(4,9)=3.68
Error 17-8=9 0.160 0.018
TOTAL 18-1=17 1.298

Interpretations:
If interaction is not significant then you may interpret main effects separately

• Present the result in the form of one way table of means for significant effects.
If interaction is significant then you cannot interpret main effects separately
• Present the result in the form of two way table of means.

8
 Interpretation of Factorial Experiments

Construct Interaction plots

Interaction Significant Yes for means of one factor at
each level of other factor
NO

Construct main effect plots

for means of significant
main effects only
 Pressure Temperature
200 215 230 Low Medium High
Means
90.367 90.683 90.183 Means
90.417 90.250 90.567

10
Example:-(2-factor in CRD with interaction*)An experimental is interested in
evaluating the relative effectiveness of three drugs in bringing about behavioral
changes in two categories,( schizophrenics and depressives). Criterion ratings are
made on each patient before and after the administration of the drugs. The
following data represents the difference between the two ratings on each of the
patient.
Drug

d1 d2 d3
Category
C1 8 4 0 10 8 6 8 6 4
C2 14 10 6 4 2 0 15 12 9

Response Variable= ratings

Levels of factor Category (c)=2
Levels of factor Drug (d)= 3
# of treatments (cxd)= 6
# replications (r)= 3
# observations (cxdxr)= 18
11
 DxC sub-table of Totals
Category Drug Drug
d1 d2 d3
d1 d2 d3 TOTAL
C1 12 24 18 54
C1 8, 4, 0 10,8, 6 8 , 6, 4 Category c2 30 6 36 72
C2 14,10,6 4, 2, 0 15, 12, 9 TOTAL 42 30 54 126

Correction Factor=(G.T)2/Total observation= (126)2/18 = 882

Total SS = (8)2+(4)2 …(9)2 – CF= 1198 - 882 = 316

(42) 2 (30) 2 (54) 2

SS D = + + − CF = 930 − 882 = 48
6 6 6
(54) 2 (72) 2
SSC = + − CF = 900 − 882 = 18
9 9
(12)2 (24)2 (18)2
SS subtable D  C = + + − CF = 1092 − 882 = 210
3 3 3
SSDxC = SSsubtable DxC- SSD – SSC=144
12
SS = 106 (By Subtraction)
 Source Of Degree of Sum of Mean Sum Fcal Ftab
Variation Freedom Squares of Squares
(DF) (SS) (MSS=SS/df)
(S.O.V)
Drug (D) 3-1 =2 48 24 2.72ns F.05(2,12)=3.89
Category (C) 2-1 =1 18 18 2.04ns F.05(1,12)=4.75
D xC 2x1 =2 144 72 8.15* F.05(2,12)=3.89
Error 17-5=12 106 8.83
TOTAL 18-1=17 316

13
 DxC TABLE OF TOTALS DxC TABLE OF MEANS
Drug Drug
d1 d2 d3
C1 d1 d2 d3
12 24 18
Category C1 4 8 6
c2 30 6 36
c2 10 2 12

Interaction Plot for Data

Fitted Means

C ategory
12
c1
c2

10

8
Mean

d1 d2 d3
Drug 14
Example:-(Three Factor Factorial in CRD) The Percentage of hardwood concentration in raw
pulp, the vat pressure, and the cooking time of the pulp are to be investigated for their effect
on the strength of the paper. Three levels of hardwood concentrations, three levels of pressure
and two cooking times are selected. A factorial experiment with two replicates is conducted in
a CRD. Response Variable= strength
% of Cooking Time
Levels of factor Time (t)= 2
Hardwood
Concentration 3hrs 4hrs Levels of factor Pressure (p)= 3
Pressure Pressure Levels of factor Concentration (c)= 3
400psi 500psi 650psi 400psi 500psi 650psi # of treatments (txpxc) = 18
196 198 213 218 211 215 # replications (r)= 3
# observations (txpxcxr) = 54
2 197

200
200

201
214

210
219

220
215

210
217

218

200 212 215 216 210 220

4 201

204
210

211
214

217
217

218
209

210
223

225

185 214 218 217 197 218

6 188

190
212

210
220

219
219

218
195

197
219

220
15
 (S.O.V) T x P x C TABLE OF TOTALS
T Cooking Time
P % of 3hrs 4hrs
C Hardwood Pressure Pressure
TxP Concentration 400psi 500psi 650psi 400psi 500psi 650psi
TxC
2 593 599 637 657 636 650
PxC
T x Px C 4 605 633 646 651 629 668
Error 6 563 636 657 654 589 657
TOTAL
PxT Table of Totals TxC Table of Totals PxC Table of Totals
Concentration
Cooking Time Cooking Time
Pressure TOTAL Concentration TOTAL Pressure 2% 4% 6%
TOTAL
3hrs 4hrs 3hrs 4hrs
400psi 1761 1962 3723 400psi 1250 1256 1217 3723
2% 1829 1943 3772
500psi 1868 1854 3722 500psi 1235 1262 1225 3722
4% 1884 1948 3832
650psi 1940 1975 3915 650psi 1287 1314 1314 3915
6% 1856 1900 3756
TOTAL 5569 5791 11360 TOTAL 3772 3832 3756 11360
TOTAL 5569 5791 11360

16
 PxT Table of Totals TxC Table of Totals PxC Table of Totals
Cooking Time Cooking Time Concentration
Pressure TOTAL Concentration TOTAL Pressure TOTAL
3hrs 4hrs 3hrs 4hrs 2% 4% 6%

400psi 1761 1962 3723 2% 1829 1943 3772 400psi 1250 1256 1217 3723

500psi 1868 1854 3722 4% 1884 1948 3832 500psi 1235 1262 1225 3722

650psi 1940 1975 3915 6% 1856 1900 3756 650psi 1287 1314 1314 3915

TOTAL 5569 5791 11360 TOTAL 5569 5791 11360 TOTAL 3772 3832 3756 11360

From TxP table of Totals

From TxC table of Totals
(5569)2 (5791)2
SST = + − CF = 912.67 (3772)2 (3832)2 (3756)2
27 27 SSC = + + − CF = 178.37
18 18 18
(3722)2 (3915) 2
SS P = + − CF = 1372.48 (1829)2 (1884)2 (1900)2
18 18 SSsubtableTxC = + + − CF = 1235.46
9 9 9
(1761) 2 (1868) 2 (1975) 2
SS subtableTxP = + + − CF = 3695.93 SSTXC = SSsubtable TxC- SST – SSC=1235.46-912.67-178.37=144.44
9 9 9
SSTXP = SSsubtable TxP- SST – SSP=3695.93-912.67-1372.48=1410.78

From PxC table of Totals

(1250)2 (1235) 2 (1314) 2

SSsubtable PxC = + + − CF = 1722.59
6 6 6
SSPXC = SSsubtable PxC- SSP – SSC=1722.59-1372.48-178.37=171.74
17
 T x P x C TABLE OF TOTALS
Cooking Time

% of 3hrs 4hrs
Hardwood Pressure Pressure
Concentration 400psi 500psi 650psi 400psi 500psi 650psi

2 593 599 637 657 636 650

4 605 633 646 651 629 668
6 563 636 657 654 589 657

From TxPxC table of Totals

(593) 2 (605) 2 (589) 2

SS subtableTxPxC = + +  − CF = 4845.85
3 3 3
SSTxPXC = SSsubtable TxPxC- SST – SSP- SSC – SSTxP - SSTxC SSPxC=655.44

18
 Source Of Variation Degree of Sum of Squares Mean Sum Fcal Ftab
(SS)
(S.O.V) Freedom of Squares
(DF) (MSS=SS/df)

Time (T) 2-1 =1 912.67 912.67 320.03* F.05(1,36)=4.12

Pressure(P) 3-1 =2 1372.48 686.24 240.63* F.05(2,36)=3.27
Concentration(C) 3-1 =2 178.37 89.18 31.27* F.05(2,36)=3.27
TxP 1x2 =2 1410.78 705.40 247.34* F.05(2,36)=3.27
TxC 1x2 =2 144.44 72.22 25.33* F.05(2,36)=3.27
PxC 2x2 =4 171.74 42.94 15.06* F.05(4,36)=2.64
TxPxC 1x2x2=4 655.44 163.86 57.46* F.05(4,36)=2.64
Error 53-17=36 102.67 2.85
TOTAL 54-1=53

As the three factor interaction is significant so main effects do not have much meaning.
Present the result in 3 way table of means for further analysis as

19
 T x P x C TABLE OF TOTALS
Cooking Time

% of 3hrs 4hrs
Hardwood Pressure Pressure
Concentration 400psi 500psi 650psi 400psi 500psi 650psi

2 593 599 637 657 636 650

4 605 633 646 651 629 668
6 563 636 657 654 589 657

T x P x C TABLE OF MEANS
Cooking Time

% of 3hrs 4hrs

Pressure Pressure
Hardwood
400psi 500psi 650psi 400psi 500psi 650psi
Concentration
2 197.67 199.67 212.33 219.00 212.00 216.67

4 201.67 211.00 215.33 217.00 209.67 222.67

6 187.67 212.00 219.00 218.00 196.33 219.00 20

 T x P x C TABLE OF MEANS
Cooking Time

% of 3hrs 4hrs
Pressure Pressure
Hardwood
400psi 500psi 650psi 400psi 500psi 650psi
Concentration
2 197.67 199.67 212.33 219.00 212.00 216.67
4 201.67 211.00 215.33 217.00 209.67 222.67
6 187.67 212.00 219.00 218.00 196.33 219.00

At time 3hrs, PxC interaction plot At time 4hrs, PxC interaction plot
Interaction Plot for data Interaction Plot for data
Fitted Means Fitted Means
220 P 225 P
400
400
500
215 500
650 220 650

210
215

205
Mean

Mean

210
200
205
195
200
190

195
2 4 6 2 4 6
C C 21
Example:- (Two Factor Factorial in RCBD)The yield of a chemical process is being studied. The two
important variables are thought to be the pressure and temperature. Three levels of each factor are selected,
however only nine runs can be made in one day. so consider the days as blocks. The yield of the experiment is
given below
Day 1 Day 2

Pressure Pressure
Temp 250 260 270 250 260 270

Low 86.3 84.0 85.8 86.1 85.2 87.3

Medium 88.5 87.3 89.0 89.4 89.9 90.3

High 89.1 90.2 91.3 91.7 93.2 93.7

Totals 791.5 806.8

22
Correction Factor=(G.T)2/Total observation= (1598.3)2/18 = 141920.16
Total SS = (86.3)2+(88.5)2 …(93.5)2 – CF= 127.069
(791.5) 2 (806.8) 2
SS Blocks = + − CF = 13.005
9 9

Source Of Degree of Sum of Squares Mean Sum Fcal Ftab

Variation Freedom (SS) of Squares
(S.O.V) (DF) (MSS=SS/df)
Blocks 1 13.005 13.005 24.48* F.05(1,8)=5.32
T 2 99.854 49.93 93.98* F.05(2,8)=4.46
P 2 5.508 2.75 5.18* F.05(2,8)=4.46
TxP 4 4.452 1.11 2.10ns F.05(4,8)=3.84
Error 8 4.250 0.53
TOTAL 17 127.069

23
24
 General ANOVA TABLES (For Factorial Experiments)
First two columns of ANOVA table for 2 and 3 factors factorial experiments

CRD
2-Factor 3-factor
S.O.V D.F S.O.V DF
A (a-1) A (a-1)
B (b-1) B (b-1)
AB (a-1)(b-1) C (c-1)
Error ab(r-1) AB (a-1)(b-1)
AC (a-1)(c-1)
BC (b-1)(c-1)
ABC (a-1)(b-1)(c-1)
Error abc (r-1)
TOTAL abr-1 TOTAL abcr-1
RCBD
2-Factor 3-factor
S.O.V D.F S.O.V DF
Block (r-1) Block (r-1)
A (a-1) A (a-1)
B (b-1) B (b-1)
AB (a-1)(b-1) C (c-1)
Error (ab-1)(r-1) AB (a-1)(b-1)
AC (a-1)(c-1)
BC (b-1)(c-1)
ABC (a-1)(b-1)(c-1)
Error (abc-1)(r-1)
TOTAL abr-1 TOTAL abcr-1

25