One-Way ANOVA : Motivation

Analysis of Variance : As QC manager, you want to Machine1 Machine2 Machine3

see if 3 filling machines have 25.40 23.40 20.00
Randomized Design different mean filling times. You
26.31 21.80 22.20
assign 15 similarly trained
workers, 5 per machine, to the 24.10 23.50 19.75
machines. At the 5% level, is 23.74 22.75 20.60
there a difference in mean 25.10 21.60 20.40
filling times ?

Randomized Design Hypotheses of One-Way ANOVA

„ Machine is called the treatment / factor / independent „ H 0 : P1 P2 " Pc
variable
„ All population means are equal
„ We say machine A is treatment 1, machine B is
„ Treatment effect is absent
treatment 2, etc.

„ We may also say there is only 1 treatment with three

levels
„ H1 : not all Pi are the same
„ Workers (experimental units) are assigned randomly to „ At least one population mean is different from the
the machines others
„ Data collected this way are analyzed by one-way „ Treatment effect is present
analysis of variance (ANOVA)
 One-Way ANOVA Assumptions No Treatment Effect
Samples are randomly and independently drawn
H 0 : P1 P2 " Pc
„

„ Populations are normally distributed

„ Populations have equal variances :

V 12 V 22 " V c2

P1 P2 P3

Treatment Effect Present Summary Measures for Each Group

Sample mean of group j :

H1 : not all Pi are the same nj
¦X
1
X j ( or X . j ) ij
nj i 1

Sample variance of group j :

2
nj
X ij  X j
S 2j ¦
i 1 n j 1
P1 P2 z P3 or
P1 z P 2 z P3
P1 P3 z P2 or
P2 P3 z P1
 Among-Groups Sum of Squares and
The Grand Mean Mean Square
c

X
n1 X 1  n2 X 2  ...  nc X c SSA : ¦ n (X j j  X )2
n1  n2  ...  nc j 1

c
§ nj · SSA
= ¦¨ n ¸ X j
M SA :
c 1
j 1 © ¹

When sample sizes are equal :

X 1  X 2  ...  X c
X Pi P j Variation Due to Differences Among Groups.
c

Within-Group Sum of Squares and

Mean Square ANOVA F Test
SSW : (n1  1) S12  (n2  1) S 22  ...  (nc  1) Sc2
Test statistic

0 0
6 6
$ :
„

)
n1 nc
¦  ...  ¦ X ic  X c
2 2
X i1  X 1
„
=
i 1 i 1
„ MSA = SSA / (c-1)
c nj
¦¦ MSW = SSW / (n-c)
2
= X ij  X j „

j 1 i 1 „ Degrees of freedom
„ numerator : c – 1
SSW „ denominator : n – c
MSW :
nc „ The F-ratio will be significantly big if the null hypothesis
is false
For c = 2, this is the pooled variance
in the pooled-variance t-test. Pj
 One-Way ANOVA Table Filling Machine Example
Machine1 Machine2 Machine3 ni 5 X1 24.93 S12 1.0648
25.40 23.40 20.00
c 3 X2 22.61 S22 0.7780
Source of Sum of Mean F 26.31 21.80 22.20
d.f. 24.10 23.50 19.75 n 15 X 3 20.59 S 2
0.9205
Variation Squares Squares Statistic 3
23.74 22.75 20.60
25.10 21.60 20.40 X 22.71
Among MSA =
SSA c–1 MSA/MSW
groups SSA/(c –1)
SSA 5 ª 24.93  22.71  22.61  22.71  20.59  22.71 º
2 2 2

MSW = ¬ ¼
Within
SSW n–c 47.164
groups SSW/(n – c)
MSA SSA / (c  1) 47.16 / 2 23.5820
Total SST n–1
MSW ( S12  S 22  S32 ) / 3 0.9211

Filling Machine ANOVA Table Filling Machine : Solution

Test Statistic:
H0: P1 = P2 = P3
Source of Sum of Mean F H1: not all equal MSA 23.5820
d.f. D = .05 F 25.6
Variation Squares Squares Statistic MSW 0.9211
df1= 2 df2 = 12

Among Decision:
47.1640 3-1=2 23.5820 25.60 Critical Value : 3.89
groups Reject at D = 0.05
Within D = 0.05 Conclusion:
11.0532 15-3=12 0.9211
groups
There is evidence that at least
Total 58.2172 15-1=14 one P i differs from the rest.
0 3.89 F
 Filling Machine : Solution by
EXCEL Filling Machine : Solution by EXCEL

ANOVA
Source of Variation SS df MS F P-value F crit
Between Groups 47.164 2 23.582 25.602 4.684E-05 3.88529
Within Groups 11.0532 12 0.9211
Total 58.2172 14

6LPXOWDQHRXV&RQILGHQFH
The Tukey Procedure ,QWHUYDOV
„ Tells which population means are significantly 7XNH\SURYHGWKDW
different from the others, e.g. ª §   ·º
MSW
„ P1 = P2 z P3 P « Pi  P j  X i  X j r QU c n c ¨¨  ¸¸ »    D
«  © ni n j ¹ »¼
„ P1 z P2 = P3 ¬
„ P1 z P2 z P3 ™Š‡”‡
„ P1 = P3 z P2 P1= P2 P3
QU c n c
‹•‘„–ƒ‹‡†ˆ”‘–Š‡•–—†‡–‹œ‡†”ƒ‰‡†‹•–”‹„—–‹‘ǡ

„ Post hoc procedure

„ Done after rejection of equal means in ANOVA
 &ULWLFDO5DQJH &ULWLFDO5DQJH
&RPSDUHPDFKLQHDQGPDFKLQH &RPSDUHPDFKLQHDQGPDFKLQH

X  X      X  X     

MSW §  ·  §   · MSW § ·  §   ·

CR QU c n  c ¨  ¸  ¨  ¸  CR QU c n  c ¨  ¸  ¨  ¸ 
 © n n ¹  © ¹  © n n ¹  © ¹

P P
6LQFH!DQGDUHGLIIHUHQW P P
6LQFH!DQGDUHGLIIHUHQW

Filling Machine : Solution by

&ULWLFDO5DQJH SPSS
&RPSDUHPDFKLQHDQGPDFKLQH

X  X    

MSW §   ·  §   ·
CR QU c n  c ¨  ¸  ¨  ¸ 
 © n  n ¹  © ¹

P P
6LQFH!DQGDUHGLIIHUHQW
Filling Machine : Solution by Filling Machine : Solution by
SPSS SPSS

Filling Machine : Solution by Filling Machine : Solution by

SPSS SPSS
ŕŦŴŵġŰŧġʼnŰŮŰŨŦůŦŪŵźġŰŧġŗŢųŪŢůŤŦŴ
ŕŊŎņ
ōŦŷŦůŦġŔŵŢŵŪŴŵŪŤ ťŧIJ ťŧij ŔŪŨį
įIJijķ ij IJij įĹĹij

łŏŐŗł
ŕŊŎņ
ŔŶŮġŰŧġŔŲŶŢųŦŴ ťŧ ŎŦŢůġŔŲŶŢųŦ Ň ŔŪŨį
ŃŦŵŸŦŦůġňųŰŶű ĵĸįIJķĵ ij ijĴįĶĹij ijĶįķıij įııı
ŘŪŵũŪůġňųŰŶűŴ IJIJįıĶĴ IJij įĺijIJ
ŕŰŵŢŭ ĶĹįijIJĸ IJĵ
 Filling Machine : Solution by
SPSS Partition of Total Sum of Squares
ŎŶŭŵŪűŭŦġńŰŮűŢųŪŴŰůŴ
ŅŦűŦůťŦůŵġŗŢųŪŢţŭŦĻġŕŊŎņ
ŕŶŬŦźġʼnŔŅ

ŎŦŢůġŅŪŧŧŦųŦůŤŦ ĺĶĦġńŰůŧŪťŦůŤŦġŊůŵŦųŷŢŭ Total Sum of

ĩŊĪġŔłŎőōņ ĩŋĪġŔłŎőōņ ĩŊĮŋĪ ŔŵťįġņųųŰų ŔŪŨį ōŰŸŦųġŃŰŶůť ŖűűŦųġŃŰŶůť
IJ ij
Ĵ
ijįĴijııī
ĵįĴĵııī
įķıķĺĺ
įķıķĺĺ
įııķ
įııı
įĸııķ
ijįĸijıķ
ĴįĺĴĺĵ
ĶįĺĶĺĵ
Squares : SST
ij IJ ĮijįĴijııī įķıķĺĺ įııķ ĮĴįĺĴĺĵ Įįĸııķ
Ĵ ijįıijııī įķıķĺĺ įıIJĶ įĵııķ ĴįķĴĺĵ
Ĵ IJ ĮĵįĴĵııī įķıķĺĺ įııı ĮĶįĺĶĺĵ Įijįĸijıķ
ij Įijįıijııī įķıķĺĺ įıIJĶ ĮĴįķĴĺĵ Įįĵııķ
īįġ ŕũŦġŮŦŢůġťŪŧŧŦųŦůŤŦġŪŴġŴŪŨůŪŧŪŤŢůŵġŢŵġŵũŦġįıĶġŭŦŷŦŭį
Among Group Within Group
= Sum of Squares : + Sum of Squares :
SSA SSW

Formula for Total Sum of Squares

c nj

SST ¦¦ ( X
j 1 i 1
ij  X )2

c = number of groups
nj = number of observations in group j
Xij = ith observation from group j
X = grand mean