FACTORIAL DESIGN FACTORIAL DESIGN
Mar 26, 2015
FACTORIAL FACTORIAL DESIGNDESIGN
•In factorial design, levels of factors are independently varied, each factor at two or more levels.•The effects that can e attributed to the factor and their interactions are assed with maximum efficiency in factorial design. So predictions based on results of an undersigned experiment will be less reliable than those which would be obtained in a factorial design.•The optimization procedure is facilitated by costruction of an equation that describes the experimental results as a function of the factorial design. Here in case of a factorial , a polynomial equation can be constructed where the coefficients in the equation are related to effects and interations of the factors.
•Now factorial design with fators at only two level is called as 2n factorial design where n is the no. of factors. these designs are simplest and often adequate to achieve the experimental objectives.
•The optimization procedure is facilitated by fitting of an empirical polynomial equation to the experimental results. The equation from for 2n factorial experiment is of the following form:
• Y= b0 + b1X1 + b2X2 + b3X3 +………+ b12X1 X2 + b13X1 X3 + b23X2 X3+……+ b123X1 X2 X3
Optimization of chromatographic Optimization of chromatographic conditions for both c8 and c18 conditions for both c8 and c18 columns carried out by a factorial columns carried out by a factorial design which evaluates temperature, design which evaluates temperature, ethanol concentration and mobile ethanol concentration and mobile phase flow rate.phase flow rate.
So design matrix would be 2So design matrix would be 233 factorial design for cfactorial design for c8 8 column.column.
NO. FACTORS LOW LEVEL HIGH LEVEL
1 TEMP (X1) 30 50
2 %ETHANOL (X2) 55 60
3FLOW RATE OF M.
PHASE (X3) 0.1 0.2
In chromatographic condition responses In chromatographic condition responses can be can be
1.1. EfficiencyEfficiency
2.2. Retention factorRetention factor
3.3. AssymetryAssymetry
4.4. Retention timeRetention time
5.5. ResolutionResolution
In this example resolution is considered In this example resolution is considered as responseas response
Experiments for a 2Experiments for a 23 3 Factorial DesignFactorial Design
NO. X1 X2 X3
1 -1 -1 -1
2 -1 1 -1
3 1 -1 -1
4 1 1 -1
5 -1 -1 1
6 -1 1 1
7 1 -1 1
8 1 1 1
Data analysis for 23 factorial design
temp %ethanol flow rate resolution/response
30 55 0.1 2.4
50 55 0.1 1.8
30 60 0.1 1.9
50 60 0.1 1.4
30 55 0.2 2.4
50 55 0.2 1.8
30 60 0.2 1.6
50 60 0.2 1.3
Coding / TransformationCoding / Transformation
The formula for transformation is The formula for transformation is X-the average of the two X-the average of the two
levelslevels one half the difference of the one half the difference of the
levelslevels
NO. X1 X2 X3 X1 X2 X1 X3 X2 X3X1 X2
X3
RESPONSE (Y)
1 -1 -1 -1 1 1 1 -1 2.4
2 -1 1 -1 -1 1 -1 1 1.8
3 1 -1 -1 -1 -1 1 1 1.9
4 1 1 -1 1 -1 -1 -1 1.4
5 -1 -1 1 1 -1 -1 1 2.4
6 -1 1 1 -1 -1 1 -1 1.8
7 1 -1 1 -1 1 -1 -1 1.6
8 1 1 1 1 1 1 1 1.3
The coefficients for polynomial The coefficients for polynomial equation are calculated as equation are calculated as
ΣΣ XY/2 XY/2nn
Where X is the value (+1 or -1) in the Where X is the value (+1 or -1) in the column appropriate for the column appropriate for the coefficient being calculated,coefficient being calculated,
Y is the response.Y is the response.
X1Y X2 Y X3Y X1X2Y X1X3Y X2X3Y X1X2X3Y Y
-2.4 -2.4 -2.4 2.4 2.4 2.4 -2.4 2.4
-1.8 1.8 -1.8 -1.8 1.8 -1.8 1.8 1.8
1.9 -1.9 -1.9 -1.9 -1.9 1.9 1.9 1.9
1.4 1.4 -1.4 1.4 -1.4 -1.4 -1.4 1.4
-2.4 -2.4 2.4 2.4 -2.4 -2.4 2.4 2.4
-1.8 1.8 1.8 -1.8 -1.8 1.8 -1.8 1.6
1.6 -1.6 1.6 -1.6 1.6 -1.6 -1.6 1.6
1.3 1.3 1.3 1.3 1.3 1.3 1.3 1.3
averageaverage
B1 b2 b3 b12 b13 b23 b123 b0
-0.275 -0.25 -0.05 0.05 -0.05 0.025 0.025 1.825
Summary output
Regression Statistics
Multiple R 1
R Square 1
Adjusted R Square 65535
Standard Error 0
Observations 8
ANOVA
Df SS MS FSignificance
F
Regression 7 1.175 0.167857 0 #NUM!
Residual 0 6.9E-31 65535
Total 7 1.175
Coefficie
ntsStandard
Error t Stat P-valueLower
95%Upper
95%
Lower 95.0%
Upper 95.0%
Intercept 1.825 0 65535 #NUM! 1.825 1.825 1.825 1.825
X1 -0.275 0 65535 #NUM! -0.275 -0.275 -0.275 -0.275
X2 -0.25 0 65535 #NUM! -0.25 -0.25 -0.25 -0.25
X3 -0.05 0 65535 #NUM! -0.05 -0.05 -0.05 -0.05
X1 X2 0.05 0 65535 #NUM! 0.05 0.05 0.05 0.05
X1 X3 -0.05 0 65535 #NUM! -0.05 -0.05 -0.05 -0.05
X2 X3 0.025 0 65535 #NUM! 0.025 0.025 0.025 0.025
X1 X2 X3 0.025 0 65535 #NUM! 0.025 0.025 0.025 0.025
RESIDUAL OUTPUT
PROBABILITY OUTPUT
ObservationPredicted
RESPONSE (Y) ResidualsStandard Residuals Percentile
RESPONSE (Y)
1 2.4 0 0 6.25 1.3
2 1.8 2.22E-16 0.797724 18.75 1.4
3 1.9 -4.4E-16 -1.59545 31.25 1.6
4 1.4 -4.4E-16 -1.59545 43.75 1.8
5 2.4 0 0 56.25 1.8
6 1.8 -2.2E-16 -0.79772 68.75 1.9
7 1.6 0 0 81.25 2.4
8 1.3 -2.2E-16 -0.79772 93.75 2.4
Multiple Linear Regression (MLR) for first reduced model after least Multiple Linear Regression (MLR) for first reduced model after least variable eliminationvariable elimination
SUMMARY OUTPUT
Regression Statistics
Multiple R 0.99787
R Square 0.995745
Adjusted R Square 0.970213
Standard Error 0.070711
Observations 8
ANOVA
df SS MS FSignificanc
e F
Regression 6 1.17 0.195 39 0.121965
Residual 1 0.005 0.005
Total 7 1.175
CoefficientsStandard
Errort Stat P-value
Lower 95%
Upper 95%
Lower 95.0%
Upper 95.0%
Intercept 1.825 0.025 73 0.00872 1.507346 2.142654 1.507346 2.142654
X1 -0.275 0.025 -11 0.057716 -0.59265 0.042654 -0.59265 0.042654
X2 -0.25 0.025 -10 0.063451 -0.56765 0.067654 -0.56765 0.067654
X3 -0.05 0.025 -2 0.295167 -0.36765 0.267654 -0.36765 0.267654
X1 X2 0.05 0.025 2 0.295167 -0.26765 0.367654 -0.26765 0.367654
X1 X3 -0.05 0.025 -2 0.295167 -0.36765 0.267654 -0.36765 0.267654
X2 X3 0.025 0.025 1 0.5 -0.29265 0.342654 -0.29265 0.342654
RESIDUAL OUTPUTPROBABILITY
OUTPUT
Observation
Predicted RESPON
SE (Y) ResidualsStandard Residuals Percentile
RESPONSE (Y)
1 2.425 -0.025 -0.93541 6.25 1.3
2 1.775 0.025 0.935414 18.75 1.4
3 1.875 0.025 0.935414 31.25 1.6
4 1.425 -0.025 -0.93541 43.75 1.8
5 2.375 0.025 0.935414 56.25 1.8
6 1.825 -0.025 -0.93541 68.75 1.9
7 1.625 -0.025 -0.93541 81.25 2.4
8 1.275 0.025 0.935414 93.75 2.4
Multiple linear regression for final reduced modelMultiple linear regression for final reduced model
Regression Statistics
Multiple R 0.969755
R Square 0.940426
Adjusted R Square
0.916596
Standard Error
0.118322
Observations 8
ANOVA
df SS MS FSignificanc
e F
Regression 2 1.105 0.552539.464
29
0.000866
Residual 5 0.07 0.014
Total 7 1.175
Normal Probability Plot
y = 0.013x + 1.1774
R2 = 0.937
0
0.5
1
1.5
2
2.5
3
0 20 40 60 80 100
Sample Percentile
RE
SP
ON
SE
(Y)
CoefficientsStd
Errort Stat P-value
Lower 95%
Upper 95%
Lower 99.0%
Upper 99.0%
Intercept 1.825 0.04183 43.625 1.19E-07 1.717465 1.932535 1.656324 1.993676
X1 -0.275 0.04183 -6.5737 0.00122 -0.38253 -0.16747 -0.44368 -0.10632
X2 -0.25 0.04183 -5.9761 0.00187 -0.35753 -0.14247 -0.41868 -0.08132
RESIDUAL OUTPUTPROBABILITY
OUTPUT
Observation
Predicted RESPON
SE (Y)Residuals
Standard Residuals
PercentileRESPON
SE (Y)
1 2.35 0.05 0.5 6.25 1.3
2 1.85 -0.05 -0.5 18.75 1.4
3 1.8 0.1 1 31.25 1.6
4 1.3 0.1 1 43.75 1.8
5 2.35 0.05 0.5 56.25 1.8
6 1.85 -0.05 -0.5 68.75 1.9
7 1.8 -0.2 -2 81.25 2.4
8 1.3 -2.2E-16 -2.2E-15 93.75 2.4
Residuals
-0.4
-0.2
0
0.2
0 1 2 3 4 5 6 7 8 9
Residuals
X1 LOW X1 HIGH
X2 LOW
1 2.4 2 1.8
5 2.4 6 1.8
2.4 1.8
X2 HIGH
3 1.9 4 1.4
7 1.6 8 1.3
1.75 1.35
prediction of interaction from graph
2.4
1.81.75
1.35
0
0.5
1
1.5
2
2.5
3
low high
value of factor
avera
ge
low
Series2
Interaction plot showing (by the parallel lines) that factors A and B do not influence each other.
Diagnostic Checking: Adjusted 2 R Rule of Thumb: Values > 0.8 typically
indicate that the regression model is a good fit.
Otherwise, a second order model is required because the linear regression is not fit for our experiment.
Final equation for this final reduced Final equation for this final reduced model will be y = 1.825-0.275*temp-model will be y = 1.825-0.275*temp-0.25*(%ethanol).0.25*(%ethanol).
Prediction from equationCoefficients of both temperature and %ethanol are having (-) negative value. So if we put lesser the value for both we will get good/ highest response / resolution.Now, batch 5 is good , so we can say that batch 5 is best which give good resolution.