Data Measurement Analysis

&
Experimental Design

Lecture 4
Model Adequacy Checking
 1
Introduction to MAC
Table
of 2
Normal Assumption Plot
Contents
3
Residual Plots
Model Adequacy Checking
 Model Adequacy Checking
Checking Error Reveal Trends of
Variance Error
 Model Adequacy Checking
Checking Error Reveal Trends of
Variance Error

1. If error appears randomly and independent then the

samples are taken randomly and independent too
2. If the error appear constantly then you know
something is wrong with the experiments (factors,
level, technical issues etc)
 Model Adequacy Checking

T test and Z test use:

• Normal Probability Plots
What statistic
analysis use ANOVA One Way use:
Model •
•
Normal Probability Plots
Residual Plots
Adequacy
Checking? ANOVA Two Ways use:
• Response Curve
• Residual Plots
• The Interaction
Video
 1
Introduction to MAC
Table
of 2
Normal Assumption Plot
Contents
3
Residual Plots
 Model Adequacy Checking for t and z tests

1. Usually used for Z- and t-test

2. Assume sample is random, drawn form
independent populations
3. Show as Normal Distribution here
4. Standard Deviation or Variances of
both populations are equal.
5. Randomise samples = satisfied
assumption
6. The plot will lie as shown, in a straight
line.
7. If a plot deviate from the line
significantly, the hypothesised model is
not appropriate.
Model Adequacy Checking for t and z tests

Formulation 1 Formulation 2
“New recipe” “Original recipe”
Model Adequacy Checking for t and z tests

How to plot Normal Assumption Plot

Step 1 Calculate the distributed percentage (%) (J – 0.5)/n

Step 2 Arrange them from least to most

Step 3 Arrange Step 2 according to percentage (%)

Step 4 Plot Step 1 (%) vs Step 3 on Normal Probability Graph

Step 5 Provide appropriate conclusion

 Step 1 - Calculate the distributed percentage (%)
using (J – 0.5)/n

• Most normal probability graph presented with this equation = (J – 0.5)/n

• Ie: there are 10 samples for both mortars formulations

• Therefore the margin of probability % is:

• (1 – 0.5)/10, (2 - 0.5)/10,…, (10 – 0.5)/10

• 5%, 15%,…, 95% distribution

• Arrange the runs y1, y2,…, yn from small to the biggest value.

% 5 15 25 35 45 55 65 75 85 95

item1
Step 2 - Arrange them from least to most

Mo 16.35 16.40 16.52 16.57 16.59 16.85 16.96 17.04 17.15 17.21

UnMo 17.50 17.63 17.75 17.86 17.90 17.96 18.00 18.15 18.22 18.25
Step 3 - Arrange Step 2 according to percentage (%)

% 5 15 25 35 45 55 65 75 85 95

Mo 16.35 16.40 16.52 16.57 16.59 16.85 16.96 17.04 17.15 17.21

UnMo 17.50 17.63 17.75 17.86 17.90 17.96 18.00 18.15 18.22 18.25

• From Step 1

• (J – 0.5)/n

• (1 – 0.5)/10, (2 - 0.5)/10,…, (10 – 0.5)/10

• 5%, 15%,…, 95% distribution

 Step 4 - Plot Step 1 (%) vs Step 3 on Normal
Probability Graph

% 5 15 25 35 45 55 65 75 85 95

Mo 16.35 16.40 16.52 16.57 16.59 16.85 16.96 17.04 17.15 17.21

UnMo 17.50 17.63 17.75 17.86 17.90 17.96 18.00 18.15 18.22 18.25
Step 5 - Provide appropriate conclusion

• Distributed normally
• No significant deviation from the
normal line
• Right side shows accumulation of
data shows the distribution is bigger
on the right side

• Distributed normally
• No significant deviation from
the normal line
• Left side shows accumulation
of data shows the distribution
is bigger on the left side
 1
Introduction to MAC
Table
of 2
Normal Assumption Plot
Contents
3
Residual Plots
Model Adequacy Checking for ANOVA One Way
Model Adequacy Checking for ANOVA One Way

How to plot Normal Assumption and Residual Plots

Calculate the Residuals and distributed percentage (%)
Step 1 (J – 0.5)/n

Step 2 Arrange them from least to most

Step 3 Arrange Step 2 according to percentage (%)

Step 4 Plot Step 1 (%) vs Step 3 on Normal Probability Graph

Step 5 Plot Residuals vs Observations/Runs

Step 6 Provide appropriate conclusion

 Step 1 - Calculate the Residuals and distributed
percentage (%) (J – 0.5)/n

• (J – 0.5)/n

• (1 – 0.5)/25, (2 - 0.5)/25,…, (25 – 0.5)/25

• 2%, 6%,…, 98% distribution

Step 2 - Arrange them from least to most
Step 3 - Arrange Step 2 according to percentage (%)
Step 4 - Plot Step 1 (%) vs Step 3 on Normal Probability Graph

1.2

Percentage
0.8

0.6

0.4

0.2

0
-5 -4 -3 -2 -1 0 1 2 3 4 5 6

Residuals
Model Adequacy Checking for ANOVA One Way
Step 5 - Plot Residuals vs Observations/Runs
Step 5 - Plot Residuals vs Observations/Runs
Step 6 - Provide appropriate conclusion

•The tendency of the graph

sloping down at the left shows
the error distribution is thinner
than normal distribution is
anticipated
•Outlier: residuals away from
normal probability plots =
distort the ANOVA
•More data are plotted on the
right side shows variance
tendency to the right of
distribution, null hypotheses
rejected
Step 6 - Provide appropriate conclusion

•The Residual vs Fitted Values

shows a horizontal-band pattern
suggests that the variance of the
residuals is somewhat constant.
•The somewhat similar size of
the groups shows the data are
randomised.
•Error are randomised as well
 1
Introduction to MAC
Table
of 2
Normal Assumption Plot
Contents
3
Residual Plots
Model Adequacy Checking for ANOVA Two Ways
Model Adequacy Checking for ANOVA Two Ways

How to plot Normal Assumption and Residual Plots

Step 1 Decide on the main factor and interaction

Calculate the Residuals and distributed percentage

Step 2 (%) (J – 0.5)/n
Arrange them from least to most

Step 3 Arrange Step 2 according to percentage (%)

Step 4 Plot Step 1 (%) vs Step 3 on Normal Probability Graph

Step 5 Plot Residuals and Interaction

Step 6 Provide appropriate conclusion

Model Adequacy Checking for ANOVA Two Ways
 Step 1 - Decide on the main factor and interaction

• Reject Null Hypothesis if F0 > F(a-1)(b-1), ab(n-1)

• Because F0.05,4,27 = 2.73, there is interaction between material types and
temperature 3.56 > 2.73
• Because F0.05,2,27 = 3.35 so the main effects interaction of material type and
temperature are also significant 7.91>3.35 and 28.97>3.35
• To assist in interpreting the results of this experiments, we will construct a graph
of the average responses at each treatment combination also known as two-
factor interaction
Step 2- Calculate the Residuals and distributed percentage (%) (J – 0.5)/n
Step 3 - Arrange them from least to most and Arrange Step 2 according to percentage (%)
Step 4 - Plot Step 1 (%) vs Step 3 on Normal Probability Graph
Step 5 - Plot Residuals and Interaction
Step 5 - Plot Residuals and Interaction
 Step 5 - Plot Residuals and Interaction
Step 5 - Plot Residuals and Interaction
Step 6 - Provide appropriate conclusion

• The Normal Probability Plot shows

a maintain regression with all
population follow the linear line.
• Although the largest residual exist,
it does not stand out somewhat
from the others
Step 6 - Provide appropriate conclusion

• It is use to predict the response at

factor levels between those actually
used in the experiment
• When at least two factors are
quantitative, we can fit a response
surface for predicting y at various
combination of the design factors.
Step 6 - Provide appropriate conclusion

• Mild tendency for the variance

of the residuals to increase as
the battery life increases
• The error variance increases as
the observations will continue
Step 6 - Provide appropriate conclusion

• Figure 5-13 and 5-14 shows

mild inequality of variance
• Treatment 15ºF and
Material Type I having
larger variance than others
• The error variance increases
as the observations will
continue
 1
Introduction to MAC
Table
of 2
Normal Assumption Plot
Contents
3
Residual Plots
Data Measurement Analysis
&
Experimental Design

Tutorial 4
Model Adequacy Checking
 Question 1

The shelf life of a carbonated beverage is of interest. Ten bottles are randomly
selected and tested, and the following results are obtained.

a) We would like to demonstrate that the mean shelf life exceeds 120 days. Set up
appropriate hypotheses for investigating this claim.
b) Test these hypotheses using α= 0.01. What are your conclusions?
c) Find the P-value for the test in part (b).
d) Construct a 99 percent confidence interval on the
mean shelf life.
e) Derive the Normal Probability Plot

FOR T DISTRIBUTION (OR Z DISTRIBUTION) ONLY NORMALITY

ASSUMPTION PLOT CAN BE DERIVED AND ENOUGH TO
SATISFY MODEL ADEQUACY CHECKING
 Question 2
The tensile strength of portland cement is being studied. Four different mixing techniques can be used
economically. The following data have been collected:
– Test the hypothesis that mixing techniques affect the strength of the cement. Use α = 0.05.
– Construct a normal probability plot of the residuals. What conclusion would you draw about
the validity of the normality assumption?
– Plot the residuals versus the predicted tensile strength. Comment on the plot
– Prepare a scatter plot of the results to aid the interpretation of the results of this experiment.

FOR ANOVA 1 WAY NORMALITY ASSUMPTION PLOT, RESIDUAL VS PREDICTED AND ONE-FACTOR
PLOT CAN BE DERIVED TO SATISFY MODEL ADEQUACY CHECKING
 Question 3
• An article in the ACI Materials Journal (Vol. 84, 1987. pp. 213-216) describes several
experiments investigating the rodding of concrete to remove entrapped air. A 3” x 6” cylinder
was used, and the number of times this rod was used is the design variable. The resulting
compressive strength of the concrete specimen is the response. The data are shown in the
following table.
– Is there any difference in compressive strength due to the rodding level?
– If there is any, please provide the post-analysis
– Use α = 0.05
– Analyze the residuals from this experiment. What conclusions can you draw about the
underlying model assumptions?

Rodding Level Compressive Strength

FOR ANOVA 1 WAY NORMALITY
ASSUMPTION PLOT, RESIDUAL VS 10 1530 1530 1440
PREDICTED AND ONE-FACTOR PLOT CAN BE 15 1610 1650 1500
DERIVED TO SATISFY MODEL ADEQUACY
20 1560 1730 1530
CHECKING
25 1500 1490 1510
 Question 4
• The yield of a chemical process is being studied. The two most important variables are thought to be
the pressure and the temperature. Three levels of each factor are selected, and a factorial experiment
with two replicates is performed. The yield data follow:
• Analyze the data and draw conclusions. Use α = 0.05.
• Prepare appropriate residual plots vs temperature and pressure and normality assumption plots
• comment on the model’s adequacy
• Prepare the interaction plot to see the interaction and best conditions to operate this process

Pressure
Temperature 200 215 230
FOR ANOVA 2 WAYS 150 90.4 90.7 90.2
NORMALITY ASSUMPTION
PLOT, RESIDUAL VS BOTH 90.2 90.6 90.4
FACTORS A AND B AND
INTERACTION PLOT CAN BE 160 90.1 90.5 89.9
DERIVED TO SATISFY MODEL 90.3 90.6 90.1
ADEQUACY CHECKING
170 90.5 90.8 90.4
90.7 90.9 90.1