Design and Analysis

of Experiments

Chapter 3

Experiments with a Single Factor:

The Analysis of Variance (ANOVA)

Dr. Tran Thanh Hung

Department of Automation Technology,
College of Engineering, Can Tho University
Email: [email protected]
 Chapter objectives

At the end of this chapter, students can:

• use ANOVA to analyze multiple
comparison experiments.
• check assumptions
• proceed when the ANOVA tells that the
mean responses differ.
• determine sample size.
Example: wafers manufacturing
Example: wafers manufacturing

(Công cụ khắc plasma)

 Example: wafers manufacturing
• An engineer is interested in investigating the relationship
between the RF power (công suất sóng) setting and the etch
rate (tốc độ khắc) for this tool. The objective of an experiment
like this is to model the relationship between etch rate and RF
power, and to specify the power setting that will give a desired
target etch rate.
• The response variable is etch rate.
• She is interested in a particular gas (C2F6) and gap (0.80 cm),
and wants to test four levels of RF power: 160W, 180W, 200W,
and 220W. She decided to test five wafers at each level of RF
power.
• The experimenter chooses 4 levels of RF power 160W, 180W,
200W, and 220W
• The experiment is replicated 5 times – runs made in random 5
order
 Example: wafers manufacturing

• Việc thay đổi công suất có

làm thay đổi tốc độ khắc?
• Mức công suất nào tối ưu
nhất?

Làm sao để phân tích dữ liệu một

cách khách quan?
Dùng t-test được không? 6
 The Analysis of Variance (ANOVA)
(Phân tích phương sai)

• In general, there will be a levels of the factor, or a treatments,

and n replicates of the experiment, run in random order (a
completely randomized design (CRD)).
• N = a x n total runs.
• Objective is to test hypotheses about the equality of the a
treatment means and to estimate them.
• Consider the Fixed Factor case: Giá trị các mức được chọn7
trước.
 The Analysis of Variance

• Models of data:
Means model : Effects model:
 i  1, 2,..., a Define: i     i
yij  i   ij , 
 j  1, 2,..., n yij     i   ij ,
i  mean of the i th factor level,   an overall mean,
 ij  experimental error.  i  i th treatment effect.

Called the one-way or single-factor analysis of variance (ANOVA) model

For hypothesis testing, the model errors are assumed to be normally and
independently distributed random variables with mean zero and variance 
2

 ij  N  0,  2  
yij  N i ,  2  8
 The Analysis of Variance
To test the equality of the a treatment means, appropriate
hypotheses are:
H 0 : 1  2    a
H1 : i   j for at least one pair (i, j )
In a effects model, we breaks a
the i th
treatment mean into:
 i a
i     i ,   i 1 ,   i  0
a i 1
 i  i treatment effect.
th

The above hypotheses can be rewritten:

H 0 : 1   2     a  0
H1 :  i  0 for at least one i 9
 The Analysis of Variance

Let:
yi. :total of the observations under the i th treatment,
yi. :average of the observations under the i th treatment,
y.. :total of all observations,
y.. :average of all observations,
n
yi.
yi.   yij , yi.  , i  1, 2,..., a
j 1 n
a n
y..
y..   yij , y..  ,
i 1 j 1 N
N  a  n :total number of observations 10
 The Analysis of Variance
• Total variability is measured by the total sum of squares(SS)
a n
SST   ( yij  y.. ) 2
i 1 j 1
• The basic ANOVA partitioning is:
a n a n

 ij ..  i. .. ij i.
( y  y
i 1 j 1
)  [( y  y2
)  ( y  y )]2

i 1 j 1
a a n a n
 n ( yi.  y.. )   ( yij  yi. )  2  ( yi.  y.. )( yij  yi. )
2 2

i 1 i 1 j 1
i 1  j 1      
0
a a n
 n ( yi.  y.. ) 2   ( yij  yi. )2  Ý nghĩa?
i 1 i 1 j 1

 SST  SSTreatments  SS E
     11
due to treatments due to error
 The Analysis of Variance

SST  SSTreatments  SS E
• A large value of SSTreatments reflects large differences in
treatment means
• A small value of SSTreatments likely indicates no
differences in treatment means
• Formal statistical hypotheses are:

H 0 : 1   2     a
H1 : At least one mean is different

12
 The Analysis of Variance
•• While sums of squares cannot be directly compared to test
 the hypothesis of equal means, mean squares can be
compared.
• A mean square is a sum of squares divided by its degrees
of freedom:
E  MS E   
2

dfTotal  dfTreatments  df Error

E  MSTreatments 
a  n  1  a  1  a (n  1)
a
SSTreatments SS E n i
MSTreatments  , MS E 
a 1 a (n  1) 2  i 1
a 1
• If the treatment means are equal, the treatment mean
square and error mean square will be (theoretically) equal.
• If treatment means differ, the treatment mean square will be
13
larger than the error mean square.
 The Analysis of Variance is
Summarized in a Table

• The reference distribution for F0 is the Fa-1, a(n-1)

distribution (randomization distribution)
• Reject the null hypothesis (equal treatment means) if
F0  F ,a 1,a ( n 1) 14
ANOVA Table
Example

15
The Reference
Distribution:

 Reject H 0

16
 Model Adequacy Checking in the
ANOVA

• Checking assumptions is important

• Normality
• Constant variance
• Independence
• Have we fit the right model?
• Later we will talk about what to do if
some of these assumptions are
violated
17
 Model Adequacy Checking in the
ANOVA (Text, p.80)
• Examination of
residuals (số dư) for
observation j in
treatment i
eij  yij  yˆij
 yij  yi.
• Residual plots are very

 ij  N  0,  2  ?
useful
• Normal probability plot
of residuals

18
Other Important Residual Plots

19
 Post-ANOVA Comparison of
Means
• The analysis of variance tests the hypothesis of equal
treatment means
• Assume that residual analysis is satisfactory
• If that hypothesis is rejected, we don’t know which
specific means are different
• Determining which specific means differ following an
ANOVA is called the multiple comparisons problem
• There are lots of ways to do this…see text, Section 3-5,
pg. 90
• We will use pairwise t-tests on means…sometimes
called Fisher’s Least Significant Difference (or Fisher’s
LSD) Method
20
 Fisher Pairwise Comparisons

Grouping Information Using the Fisher LSD Method and 95% Confidence
RF
power N Mean Grouping
220 5 707.00 A
200 5 625.40 B
180 5 587.40 C
160 5 551.20 D
Means that do not share a letter are significantly different.

Fisher Individual Tests for Differences of Means

Difference Difference SE of Adjusted
of Levels of Means Difference 95% CI T-Value P-Value
180 - 160 36.2 11.6 ( 11.7, 60.7) 3.13 0.006
200 - 160 74.2 11.6 ( 49.7, 98.7) 6.42 0.000
220 - 160 155.8 11.6 (131.3, 180.3) 13.49 0.000
200 - 180 38.0 11.6 ( 13.5, 62.5) 3.29 0.005
220 - 180 119.6 11.6 ( 95.1, 144.1) 10.35 0.000
220 - 200 81.6 11.6 ( 57.1, 106.1) 7.06 0.000
Simultaneous confidence level = 81.11%
21
The Regression Model

22
 Sample Size Determination

• Can choose the sample size to detect a specific

difference in means and achieve desired values of
type I and type II errors
• Type I error – reject H0 when it is true (  )
• Type II error – fail to reject H0 when it is false (  )
• Power = 1 - 
• Operating characteristic curves plot  against
a parameter  where
a
n  i2
2  i 1

a 2 23
 Sample Size Determination
Fixed Effects Case: use of OC Curves
• The OC curves for the fixed effects model are in
the Appendix, Table V, pg. 615
• A very common way to use these charts is to
define a difference in two means D of interest,
then the minimum value of  is
2

2
nD
2 
2a 2
• Typically work in term of the ratio of D /  and
try values of n until the desired power is
achieved
• Minitab can perform power and sample size
calculations

24
 Thực hành chương 3

• Bài 1: Dùng Minitab phân tích kết quả thí

nghiệm trong ví dụ khắc wafer.

• Bài 2: Dùng cùng một loại giấy, xếp cùng

một loại máy bay với 3 độ rộng cánh
khác nhau. Phóng thử mỗi loại 10 lần,
ghi lại độ xa mỗi lần phóng. Dùng Minitab
phân tích kết quả. Ước lượng mô hình
toán liên hệ giữa độ rộng cánh và độ xa.