Vous êtes sur la page 1sur 37

Asst. Prof. Florence C. Navidad, RMT, RN, M.Ed.

1  
Objec&ves  

At  the  end  of  the  topic,  the  students  are  expected  to:  
•  Be  familiar  with  the  use  of  ANOVA  in  hypothesis  
tes&ng;  
•  Differen&ate  the  three  types  of  varia&ons  that  are  
considered  in  Anova  tests;  
•  Dis&nguish  between  One-­‐way  and  Two-­‐way  Anova  
designs  
•  Perform  a  Post  Anova  test  aGer  each  hypothesis  
tested  using  the  Tukey’s  HSD.  

2  
Different Forms of ANOVA  

•  One-­‐way  ANOVA  (one-­‐factorial  designs)  


•  One-­‐Way  Between-­‐Groups  (independent  groups)  ANOVA  
•  One-­‐Way  Within-­‐Subjects  (Repeated  Measures)  ANOVA  
•  Two  -­‐way  ANOVA  (two-­‐factorial  designs)  
•  Two-­‐Way  Between-­‐Groups  (Independent  groups)  ANOVA  
•  Post  –Hoc  Analysis  

3  
4  
One-­‐Way  or  One-­‐Factor  Anova  
[One-­‐Way  Between-­‐Groups  (Independent  groups)]  

•  Used  when  the  experiment  has  more  than  two  


levels  of  the  independent  variable.  
•  results  stop  when  decision  is  the  acceptance  of  the  
null  hypothesis.  
•  when  decision  is  the  rejec&on  of  null  hypothesis,  
search  con&nues  which  pair  accounts  for  the  
difference.  

5  
Steps  in  Conduc&ng  One-­‐Way  Anova              

1.  Formulate the null and alternative hypothesis.


2.  Choose a level of significance.
3.  Choose the appropriate test statistic and
Compute the value of the statistical test
4.  Establish the critical region
5.  Make a decision
Decision Rule: Do not accept Ho if the test
statistic value lies inside the critical region,
otherwise accept Ho
6. State the appropriate conclusion.
6  
ANOVA  TABLE  

Source  of   Sum  of   dF Mean  Squares F  Sta2s2c


Varia2on Squares
Between   SSb k-­‐1 MSb  =  SSb/k-­‐1 MSb/MSw
Groups
Within   SSw N-­‐k MSw    =  SSw/(N-­‐k)
Groups
Total SSt N-­‐1  

7  
Example  1  

•  Weight  gain  (oz)  of  20  


baby  rats  fed  four   A   B   C   D  
different  brands  of   1   7   9   8  
cereal.   1   7   6   6  
1   7   5   4  
•  Is  there  a  significant  
1   7   3   1  
difference  in  weight   1   7   2   1  
gain  among  the  rats  
that  were  fed  the  
different  brands  of  
cereal?  

8  
Steps  in  Hypothesis  Tes&ng  
H0:  there  no  significant  difference  in  weight  gain  among  the  rats  
that  were  fed  the  different  brands  of  cereal  
H1:  there  is  a  significant  difference  in  weight  gain  among  the  rats  
that  were  fed  the  different  brands  of  cereal  

 α  =  0.05  
One-­‐way  ANOVA:  SPSS  
Cri&cal  region:  F  (3,16)  3.24    &  F-­‐test:    7.353  
P-­‐value:    0.003  <  0.05    
Decision:  do  not  cccept  null  hypothesis  and  accept  alterna&ve  
hypothesis        

Conclusion:  This  means  that  at  least  one  of  the  means  is  
significantly  different  from  the  others.       9  
POST  HOC  ANALYSIS  

•  When  the  null  hypothesis  is  rejected,  it  may  be  


desirable  to  find  which  mean(s)  is  (are)  different,  
  and  at  what  ranking  order.  
•  Three  most  commonly  used  sta&s&cal  inference  
procedures,  geared  at  doing  this,  are  presented:  
–  Fisher’s  least  significant  difference  (LSD)  
method  
–  Tukey  HSD  test  (Tukey)  
–  Student-­‐Newman-­‐Keuls  (S-­‐N-­‐K)  
10  
Mul&ple  comparison  
Example  1  

Pair     P-­‐value   decision   Results  


A  –  B   0.002   p  <  α   Significant  
A  –  C   0.033   p  <  α   Significant  
A  –  D   0.139   p  >  α   Not  significant  
B  –  C   0.442   p  >  α   Not  significant  
B  –  D   0.139   p  >  α   Not  significant  
C  –  D   0.868   p  >  α   Not  significant  
 

11  
EXAMPLE  2  
A  medical  equipments  retailer  is  
selling  4  brands  of  stethoscope.     A   B   C   D  
7   9   2   4  
The  owner  interested  if  there  is  
3   8   3   5  
a  significant  difference  in  the  
5   8   4   7  
average  sales  of  the  four  
6   7   5   8  
brands  of  stethoscope  for  one   9   6   6   3  
week.    Perform  the  f-­‐test  and   4   9   4   4  
test  the  hypothesis  at  0.05   3   10   2   5  
level  of  significance  that  the  
average  sales  of  the  four  
brands  of  stethoscope  are  
equal.  
12  
Steps  in  Hypothesis  Tes&ng  
H0    :    There  is  no  significant  difference  in  the  average  sales  of  the  
four  brands  of  stethoscope.  
H1    :    There  is  a  significant  difference  in  the  average  sales              
of  the  four  brands  of  stethoscope.  
 
 α  =  0.05  
One-­‐way  ANOVA:  SPSS  
Cri&cal  region:  F  (3,24)  3.01    &  F-­‐test:    7.969  
P-­‐value:    0.001  <  0.05    
Decision:  do  not  accept  null  hypothesis  and  accept  alterna&ve  
hypothesis        

Conclusion:  Therefore,  there  is  a  significant  difference  in  the  


average  sales  of  the  4  brands  of  stethoscope.   13  
Mul&ple  comparison  
Example  2  

Pair     P-­‐value   decision   Results  


A  –  B   0.025   p  <  α   Significant  
A  –  C   0.350   P  >  α   not  Significant  
A  –  D   0.999   p  >  α   Not  significant  
B  –  C   0.000   P  <  α    significant  
B  –  D   0.018   P  <  α    significant  
C  –  D   0.432   p  >  α   Not  significant  
 

14  
Example  3  
Three  brands  of  reducing  pills  were  tried  on  a  sample  of  8  female  
adults.    The  data  is  reflected  on  the  table  below  in  terms  of  
weight  loss  (lb).  Test  the  hypothesis  that  there  is  no  significant  
difference  in  the  average  weight  loss  (in  lb)  among  the  three  
groups  of  respondents  using  the  3  brands  of  reducing  pill  at  the  
0.01  level.  
Respondents   Brand  A   Brand  B   Brand  C  
no.  
1   4.5   3.2   3.0  
2   4.1   3.0   2.8  
3   3.6   3.8   3.2  
4   5.3   3.9   3.6  
5   4.8   4.2   3.5  
6   2.7   3.1   3.5  
7   4.3   4.0   2.9  
8   3.8   3.3   3.6   15  
One-­‐Way  or  One-­‐Factor  Anova  
[One-­‐Way  Within-­‐Subjects  (Repeated  Measures)]  

•  Tes2ng  the  null  hypothesis  that,  on  the  


average,  the  scores  of  subjects  remain  the  
same  during  each  of  the  different  condi2ons  
(the  baseline  condi2on  and  the  2,3,  or  4  
different  condi2ons)  

16  
Example  1  

•  The  facilitator  of  a  stress  management  therapy  group  


conducted  a  study  to  determine  the  most  effec2ve  
relaxa2on  technique/s  for  stress  reduc2on.  The  10  
members  of  his  stress  management  group  
par2cipated  in  the  study.    The  heart  rate  of  each  
par2cipant  was  monitored  during  each  of  five  
condi2ons.    Each  par2cipant  experienced  all  five  
condi2ons  during  the  same  session  to  control  for  
varia2ons  in  the  amount  of  stress  experienced  from  
day  to  day.  

17  
The  mean  heart  rate  (beats  per  minute)  for  each  subject  during  
each  condi2on  is  as  follows:  

Sub   Baseline   Medita& Comedy   Nature   Music  


on  
1   85   70   75   71   74  
2   79   69   73   70   72  
3   91   82   87   83   86  
4   93   80   85   79   84  
5   92   80   86   81   87  
6   87   79   83   80   81  
7   84   72   77   73   76  
8   78   69   74   71   73  
9   79   69   73   70   72  
10   80   71   74   72   73   18  
Steps  in  Hypothesis  Tes&ng  
H0    :    There  is  no  significant  difference  in  the  mean  heart  rates  of  
subjects  during  each  of  the  five  condi&ons.  
H1    :  There  is  a  significant  difference  in  the  mean  heart  rates  of  
subjects  during  each  of  the  five  condi&ons.  
   α  =  0.05  
One-­‐way  ANOVA:  SPSS  
Cri&cal  region:   P-­‐value:    0.000  <  0.05    
NOTE:  Read  test  within-­‐subjects  effect  for  the  analysis  or  
Mul&variate  tests  
Decision:  do  not  accept  null  hypothesis  and  accept  alterna&ve  
hypothesis        
Conclusion:  There  is  a  significant  difference  in  the  mean  heart  
rates  of  subjects  during  each  of  the  five  condi&ons.   19  
Example  2  

•  The  data  were  collected  in  a  study  inves&ga&ng  


factors  associated  with  the  risk  of  developing  high  
blood  pressure,  or  hypertension.    This  test  
determines  if  there  is  evidence  that  the  diastolic  
blood  pressure  changes  significantly  during  three  
different  condi&ons:  res&ng,  doing  mental  
arithme&c,  and  immersing  a  hand  in  a  cold  water.  

Ho:    blood  pressure  does  not  change  during  the  


stressors  
20  
Dbprest   Dbpma   Dbpcp  

60.75   81.33   65.00  


76.38   82.67   80.00  
65.38   90.33   90.00  
72.13   82.33   89.00  
68.75   75.33   70.00  
73.50   82.33   70.00  
57.13   68.00   62.00  
65.63   78.67   73.00  
67.38   81.67   66.00  
56.50   58.67   65.00  

21  
RANDOMIZED  BLOCKS    
(TWO-­‐WAY)  ANALYSIS  OF  VARIANCE  
[Two-­‐Way  Between-­‐Groups  (Independent  
Groups)  

22  
•  Two  independent  variables  at  the  same  &me  
•  The  purpose  of  designing  a  randomized  block  
experiment  is  to  reduce  the  within-­‐treatments  
varia4on.  
•  Determine  whether  there  is  significant  
difference  in  the  mean  score  from  the  
different  methods  of  treatments.  (columns)  
•  Determine  whether  there  is  a  difference  in  the  
mean  scores  obtained  by  the  sample  
(students).  (rows)    
•  Major  difference:  We  now  have  3  separate  tests,  
and  three  separate  F  values  
23  
Randomized  Blocks  
Block  all  the  observa&ons  with  some    
commonality  across  treatments  

Treatment  4  
Treatment  3  

Treatment  2  
Treatment  1  

Block3   Block2   Block  1  


24  
Anova  Table  for  the  Randomized  
Complete  Block  Design  

Source  of   Sum  of   dF Mean  Squares F  Sta&s&c


Varia&on Squares
Treatments   SStr k-­‐1 SStr/k-­‐1 MS(SStr)/
(column)   MS(SSr)
Blocks  (rows) SSb n-­‐1 SSb/n  -­‐  1
Residual   SSr (k  –  1)   SSr/(k  –  1)(n  –  1)
(error)   (n  –  1)
Total SSt Kn  -­‐  1

25  
 
Three  dis&nct  hypothesis  tests  
 
•  Test  for  the  main  effects  
–  The  mean  difference  between  levels  of  the  first  
factor.  
–  The  mean  difference  between  level  of  the  second  
factor  
•  Test  for  the  interac&on  
–  Any  other  mean  differences  that  may  result  from  
the  unique  combina&on  of  the  two  factors.  

26  
EXAMPLE  1  

Forty-­‐five  sta&s&cs  students  were  randomly  


assigned  to  one  of  three  instructors  and  to  
one  of  the  three  methods  of  teaching.    
Achievement  was  measured  on  a  test  
administered  at  the  end  of  the  term.    Use  the  
two-­‐way  ANOVA  at  0.05  level  of  significance  
to  test  the  following  hypothesis.  

27  
TWO-­‐FACTOR  ANOVA  with  Significant  Interac&on    

TEACHER  FACTOR  
    A   B   C  
Method  of  Teaching  1   40   50   40  
41   50   41  
40   48   40  
39   48   38  
38   45   38  
Method  of  Teaching  2   40   45   50  
41   42   46  
39   42   43  
38   41   43  
38   40   42  
Method  of  Teaching  3   40   40   40  
43   45   41  
41   44   41  
39   44   39  
38   43   38   28  
Problem  

•  Is  there  a  significant  difference  in  the  performance  


of  students  under  the  three  different  teachers?  
•  Is  there  a  significant  difference  in  the  performance  
of  students  under  the  three  different  methods  of  
teaching?  
•  Is  there  a  significant  difference  in  the  interac&on  of  
the  three  groups  of  students  under  three  different  
methods  of  teaching  and  instructors?  

29  
Hypothesis  
1.  H0    :    There  is  no  significant  difference  in  the  
performance  of  the  three  groups  of  students  under  
three  different  instructors.  
2.      H0    :    There  is  no  significant  difference  in  the  
performance  of  the  three  groups  of  students  under  
three  different  methods  of  teaching.  
3.      H0    :  There  is  no  significant  difference  in  the  
interac&on  of  the  three  groups  of  students  under  
three  different  methods  of  teaching  and  instructors.  
 

30  
Descrip&ve  
A   B   C   total  
Method  1   39.6   28.2   39.4   42.4  
Method  2   39.2   42   44.8   42  
Method  3   40.2   43.2   39.8   41.07  
total   39.67   44.47   41.3  

31  
Conclusion  
1.  Computed  F-­‐value  (column).    Accept  null  hypothesis.    
Therefore,  there  is  no  significant  differences  in  the  
performance  of  the  three  groups  of  students  under  three  
different  instructors.      
2.  Computed  F-­‐value  (row).    Accept  null  hypothesis.    
Therefore,  there  is  no  significant  differences  in  the  
performance  of  the  students  under  the  three  different  
methods  of  teaching  .  
3.  Computed  F-­‐value  (column-­‐row).    Reject  null  hypothesis.    
Therefore,  there  is  a  significant  difference  in  the  
interac&on  of  the  three  groups  of  students  under  three  
different  methods  of  teaching  and  instructors.  

32  
Example  2  

A  male  psychology  student  decides  to  conduct  a  study  


to  determine  the  effec&veness  of  various  approaches  
to  mee&ng  someone.  The  three  approaches  he  wants  
to  inves&gate  are  as  follows:  causal  conversa&on  
approach,  humor  approach,  and  pick-­‐up  line  
approach.  He  also  decided  to  include  aurac&veness  as  
a  2nd  independent  variable.    The  measure  of  success  
for  each  approach  is  the  length  of  &me  in  minutes.    
The  results  are  as  follows  using  0.05  level  of  
confidence.  

33  
Subject   aurac&ve approach   &me  
nss  
1   1   1   43  
2   1   1   35  
3   1   1   52  
4   1   1   61  
5   1   2   45  
6   1   2   52  
7   2   2   30  
8   2   2   45  
9   2   3   1  
10   2   3   2  
11   2   3   3  
12   2   3   4  
34  
Hypothesis  

1.  Ho:    there  is  no  significant  difference  in  the  main  
effect  of  aurac&veness.  
2.  H0:    there  is  no  significant  difference  in  the  main  
effect  approach  type;  
3.  H0:  there  is  no  significant  difference  in  the  
interac&on  of  aurac&veness  and  approach  type  

35  
Conclusion  

1.   there  is  no  significant  difference  in  the  main  


effect  of  aurac&veness.  
2.   there  is  a  significant  difference  in  the  main  effect  
approach  type;  
3.   there  is  a  significant  difference  in  the  interac&on  
of  aurac&veness  and  approach  type  

36  
The  end  

“It is possible to fail in many ways…


While to succeed is possible only in one
way.”
- Aristotle

37  

Vous aimerez peut-être aussi