Вы находитесь на странице: 1из 11

International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

1

STATISTICS FOR BUSINESS
CHAPTER 09
ANALYSIS OF VARIANCE

ANOVA is statistical method for determining the existence of differences among several
population means.
The required assumptions of ANOVA
1. We assume independent random sampling from each of the r populations.
2. We assume that the r populations under study are normally distributed, with means
p, that may or may not be equal, but with equal variance o
2


PART I
ANOVA TABLE AND THE COMPUTATIONS OF ANOVA
ANOVA TABLE
Source of Variation
Sum of Squares
(SS)
D
Mean Squares
(MS)
F. ratio
(F
T
)
Treatment
(TR)
SSIR r 1 HSIR =
SSIR
r 1
F
1
=
HSIR
HSE

Error
(F)
SSE n r
HSE =
SSE
n r


Total
(T)
SSI n 1




International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

2

Computations of x
|
and x
The mean of sample i (i =1,,r): x
|
=_x
|j
n
|
j=1
_/ n
|

The grand mean (the mean of all the data points): x =_x
|j
n
|
j=1
r
j=1
_/ n

Note:
- x
]
is the particular data point in position ] within the sample from population i.
- The subscript i denotes the population, or treatment, and runs from 1 to r.
- The subscript ] denotes the data point within the sample from population i; thus, ] runs
from 1 to n

.

The Sum-of-Squares Principle
The sum of squares for treatment: SSTR =n
|
(x
|
x)
2
r
|=1

The sum of squares for error: SSF =(x
|j
x
|
)
2
n
|
j=1
r
|=1

The sum-of-squares total: SST =SSTR +SSF =(x
|j
x)
2
n
|
j=1
r
|=1


The Degree of Freedom
The number of degrees of freedom associated with SSTR:
d(treatment) =r 1
The number of degrees of freedom associated with SSE:
d(errur) =n r

International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

3

The number of degrees of freedom associated with SST:
d(tuta|) =n 1

The Mean Squares
Mean Square Treatment MSTR =
SSTR
r 1

Mean Square Error MSF =
SSF
n r


The F statistic
The test statistic value (of the hypothesis testing for ANOVA):
F
T
=F rat|u =
MSTR
MSF

Under the assumptions of ANOVA, the ratio HSIR HSE possesses an F distribution with r 1
degrees of freedom for the numerator and n r degree of freedom for the denominator when null
hypothesis is true.

CALCULATOR INSTRUCTION
(Computing SSTR, SSE, and SST by using pocket calculation)

The following data represent the result of running an experiment under four different
levels A, B, C, and D of a single factor.




International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

4

A B C D
1 18 17.75 17.92 18.01
2 17.98 18 18.01 17.94
3 18.2 17.77 17.88 18.23
4 18 18.01 18.3 18.2
5 17.99 18.01 18.22 18
6 18.1 18.12 18.56 17.84
7 17.9 18.2 18.1 18.11

Using the calculator to compute the sum of square for treatment
SSTR =n
|
(x
|
x)
2
r
|=1

Firstly, we calculate the sample mean of each sample A, B, C, and D using the following
method of each treatment.
Step 1: Set up STATISTIC back ground
Press MODE +3: STAT
Step 2: Input the data
Press 1: 1 VAR
Step 3: Get the result
Press SHIFT +1 [STAT] +5: VAR +2: x
A B C D
Sample Mean
x


18.024 17.980 18.141 18.470



International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

5

Next, we calculate the value of standard deviation of SSTR by following these steps based
on FREQUENCY method.
Step 1: Set up FREQUENCY method
Press SHIFT +MODE [SET UP]
REPLAY +4: STAT
[Frequency ?] 1: ON
Step 2: Set up STATISTIC background
Press MODE +3: STAT
Step 3: Input the data
Press 1: 1 VAR
MODEL EXAMPLE
x FREQ x FREQ
1 x
1
n
1
1 18.024 7
2 x
2
n
2
2 17.980 7
3 x
3
n
3
3 18.141 7
4 18.047 7
R x



Note:
- r : number of samples or treatments
- n : number of observations of each treatment
- x : sample mean of each treatment

Step 04: Get the result
Press SHIFT +1 [STAT] +5 : VAR +3 : xon
The result we can get is SD =.

International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

6

Finally, we calculate SSTR based on the following equation
SSIR =n

(x

x )
2

=1

Step 01: SD =.
Step 02: FAR =[SD]
2
=(.)
2
=.3
n 1 =28 1 =27
Step 03: SSTR =(FAR)(n 1) =(.3)(27) =.972

Note: n : number of observations of r treatments

Using the calculator to compute the sum of square for error
SSF =(x
|j
x
|
)
2
n
|
j=1
r
|=1

Firstly, we calculate the sample standard deviation of each sample A, B, C, and D by using
the following method for each treatment.
Step 01: Set up STATISTIC background
Press MODE +3 : STAT
Step 02: Input the data
Press 1 : 1 VAR
Step 03: Get the result
Press SHIFT +1 [STAT] +5 : VAR +4 : xon 1


International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

7

A B C D
Sample Standard
Deviation
s


0.0969 0.1669 0.2391 0.1407

Next, we calculate SSE based on the following equation
SSE =(x
]
x

)
2
n
i
]=1

=1

(Hint: Establishing a table for computations)
MODEL
METHOD Treatment 1 Treatment 2 Treatment r
Step 01 S s
1
s
2
s


Step 02
IAR =[S]
2
s
1
2
s
2
2
s

2

n

1 n
1
1 n
2
1 n

1
Step 03 (IAR)(n

1) (s
1
2
)(n
1
1) (s
2
2
)(n
2
1) (s

2
)(n

1)
Step 04 SSE =SSE

=1
[(s
1
2
)(n
1
1)] +[(s
2
2
)(n
2
1)] ++[(s

2
)(n

1)]

EXAMPLE
METHOD Treatment 1 Treatment 2 Treatment 3 Treatment 4
Step 01 S 0.0969 0.1669 0.2391 0.1407
Step 02
IAR =[S]
2
0.0091 0.0279 0.0571 0.0198
n

1 6 6 6 6
Step 03 (IAR)(n

1) 0.0564 0.1672 0.3429 0.1187


Step 04 SSE =SSE

=1
0.0564 +0.1672 +0.3429 +0.1187 =0.6852



International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

8

Using the calculator to compute the sum-of-square total
SST =SSTR +SSF =(x
|j
x)
2
n
|
j=1
r
|=1

Firstly, we treat n

observations of r treatments as of a large sample.


Next, we use the calculator to compute the value of standard deviation of the large sample
by following these steps.
Step 01: Set up STATISTIC background
Press MODE +3 : STAT
Step 02: Input the data
Press 1 : 1 VAR
MODEL EXAMPLE
x x
1 1
st
observation 1 18
2 2
nd
observation 2 17.98
3 3
rd
observation 3 18.2
4 4
th
observation 4 18

n

last observation 49 18.11



Step 04: Get the result
Press SHIFT +1 [STAT] +5 : VAR +4 : xon 1
The result we can get is SD =.173

Finally, we calculate SST based on the following equation
SSI =SSIR +SSE =(x
]
x )
2
n
i
]=1

=1

International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

9

Step 01: SD =.173
Step 02: FAR =[SD]
2
=(.173)
2
=.29
n 1 =28 1 =27
Step 03: SST =(FAR)(n 1) =(.29)(27) =.782

* We just need to calculate two of three values, SSTR, SSE, and SST. Then with the
equation SST =SSTR +SSF, we can find the remaining value.

PART II
HYPOTHESIS TESTING FOR ANOVA

HYPOTHESIS TESTING PROCESS:
Step 01: We assume the independent random samples (or treatments) and the
normally distributed populations.
Step 02: Determine the null and alternative hypotheses
E
0
: p
1
=p
2
=p
3
= =p


E
1
: Not oll p

(i =1,2,3,,r) orc cquol


Step 03: Construct the ANOVA table
Source of Variation
Sum of Squares
(SS)
D
Mean Squares
(MS)
F. ratio
(F
T
)
Treatment
(TR)
SSIR r 1 HSIR =
SSIR
r 1
F
1
=
HSIR
HSE

Error
(F)
SSE n r HSE =
SSE
n r


Total
(T)
SSI n 1
International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

10

Step 04: Compute the test statistic value (F
1
) (based on the ANOVA table) and the
critical value (F
C
) (based on the level of significance)
The test statistic value:
F
T
=F rat|u =
MSTR
MSF

The critical value with the level of significance (o)
(Appendix C, Table 5 or Table 5A)
F
C
=F
(u,r-1,n-r)


Step 05: Compare the test statistic value with the critical value and make a decision
At the level of significance (o),
+Situation 1: We can reject E
0
when F
1
>F
C

+Situation 2: We cannot reject E
0
when F
1
<F
C


Sample:
PROBLEM
(ANOVA)

Three checkout lines at a supermarket use three different scanner systems that read
the UPC symbols on products and nd the prices. The store manager suspects that the
three scanner systems have different efficiencies and wants to check their speeds. He
measures at randomly selected times the speed of each system in number of items
scanned per minute. The measurements are given in the table below. Assume normal
distribution with equal variance for the three systems.

Scan 1 Scan 2 Scan 3
16 13 18
15 18 19
12 13 15
15 15 14
16 18 19
15 14 16
15 15 17
International University IU

TA [SWC]
S
t
a
t
i
s
t
i
c
s

f
o
r

B
u
s
i
n
e
s
s

|

C
h
a
p
t
e
r

0
9
:

A
n
a
l
y
s
i
s

o
f

V
a
r
i
a
n
c
e

11

14 15 14
12 14 15
14 16 17
SOLUTION: (ANOVA: Single Factor)

STEP 01: We assume that the three treatments (or three samples) are
independent of each other and three populations are normally
distributed

STEP 02: E
0
: p
1
=p
2
=p
3

E
1
: Not oll p

(i =1,2,3) orc cquol



STEP 03: SUMMARY

Group Count Sum Average Variance
Scan 1 10 144 14.4 2.044
Scan 2 10 151 15.1 3.211
Scan 3 10 164 16.4 3.600

ANOVA Table

Source of Variation SS Df MS F ratio
Treatment 20.6 2 10.300 3.489
Error 79.7 27 2.952
Total 100.3 29

STEP 04: The test statistic value:
F
1
=F rotio =3.489

At o =0.05, the critical value:
F
C
=F
(2,27,0.05)
=3.354

STEP 05: Thus, at 0.05 level of significance, we can reject the null
hypothesis since F
1
>F
C
. It means that based on the ANOVA
table and the hypothesis testing we have sufficient evidence to
prove that not all three scanner systems have the same average
number scanned per minute.

Вам также может понравиться