50% found this document useful (2 votes)
540 views38 pages

Analysis of Variance PPT Powerpoint Presentation3260 PDF

The ANOVA test compares the means of 3 groups (control, treatment 1, treatment 2) on a mobility score. It tests the null hypothesis that all group means are equal versus the alternative that at least two means differ. The ANOVA F-statistic of 5.73 exceeds the critical value from the F distribution with 2 and 15 degrees of freedom at the 0.05 significance level. Therefore, the null hypothesis is rejected, indicating there are significant differences in mobility scores between at least two of the groups.

Uploaded by

Hamid Husain
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
50% found this document useful (2 votes)
540 views38 pages

Analysis of Variance PPT Powerpoint Presentation3260 PDF

The ANOVA test compares the means of 3 groups (control, treatment 1, treatment 2) on a mobility score. It tests the null hypothesis that all group means are equal versus the alternative that at least two means differ. The ANOVA F-statistic of 5.73 exceeds the critical value from the F distribution with 2 and 15 degrees of freedom at the 0.05 significance level. Therefore, the null hypothesis is rejected, indicating there are significant differences in mobility scores between at least two of the groups.

Uploaded by

Hamid Husain
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 38

ANOVA

(Analysis of Variance)
ANOVA: Introduction
• Many studies involve comparisons between more
than two groups of subjects.
• If the outcome is categorical (count) data, a Chi-
square test for a larger than 2 X 2 table can be used to
compare proportions between groups.
• If the outcome is numerical, ANOVA can be used to
compare the means between groups.
• ANOVA is an abbreviation for the full name of the
method: ANalysis Of Variance
– Invented by R.A. Fisher in the 1920’s
Why ANOVA instead of multiple
t-tests?
• If you are comparing means between more than two
groups, why not just do several two sample t-tests to
compare the mean from one group with the mean from
each of the other groups?
– Before ANOVA, this was the only option available to
compare means between more than two groups.
• The problem with the multiple t-tests approach is that
as the number of groups increases, the number of two
sample t-tests also increases.
• As the number of tests increases the probability of
making a Type I error also increases.
ANOVA: a single test for
multiple comparisons
• The advantage of using ANOVA over multiple
t-tests is that ANOVA will identify if any two of
the group means are significantly different with a
single test.
• If the significance level is set at 0.05, the
probability of a Type I error for ANOVA = 0.05
regardless of the number of groups being
compared.
• If the ANOVA F-test is significant, further
comparisons can be done to determine which
groups have significantly different means.
ANOVA Hypotheses
• The Null hypothesis for ANOVA is that the means for
all groups are equal:
H o : µ1 = µ 2 = µ3 = .... = µ k
• The Alternative hypothesis for ANOVA is that at least
two of the means are not equal.
• The test statistic for ANOVA is the ANOVA
F-statistic.
Analysis of Variance
• ANOVA is used to compare means between three or
more groups, so why is it called Analysis of
VARIANCE?
• The ANOVA F-test is a comparison of the average
variability between groups to the average variability
within groups.
– The variability within each group is a measure of the spread of
the data within each of the groups.
– The variability between groups is a measure of the spread of
the group means around the overall mean for all groups
combined.
– F = average variability between groups
average variability within groups
ANOVA: F-statistic
• If variability between groups is large relative to
the variability within groups, the F-statistic will be
large.
• If variability between groups is similar or smaller
than variability within groups, the F-statistic will
be small.
• If the F-statistic is large enough, the null
hypothesis that all means are equal is rejected.
Illustration of small F-statistic

Within
Group
Group 1
mean
Group 2
mean
Overall mean
Group 3
mean

Source: Introduction to the Practice of Statistics, Moore and McCabe


Illustration of large F-statistic

Group 1 Within group


mean

Group 2
mean
Overall
mean Group 3
mean

Source: Introduction to the Practice of Statistics, Moore and McCabe


Total variation in two parts
• The difference of each observation from the
overall mean can be divided into two parts:
– The difference between the observation and the
group mean
– The difference between the group mean and the
overall mean
• ANOVA makes use of this partitioned
variability.
Partitioned Variability
• Consider the following VERY small data set
Mean for each group
– Group 1: 12 16 18 _______
– Group 2: 10 12 16 _______
– Group 3: 8 9 10 12 _______
– Overall mean = _______
• Subject 3 in group 1 has value 18
– Difference between subject 3 and overall mean:
18 - _____ = _____
– This can also be expressed as the difference between
subject 3 and the mean for group 1 + the difference
between the mean for group 1 and the overall mean:
(18 - ___ ) + ( ____ - ____ ) = ______
SST = SSW + SSB
• This partitioned relationship is also true for the
squared differences:
– The variability between each observation and the
overall (or grand) mean is measured by the ‘sum of
squares total’ (SST)
– The variability within groups is measured by the ‘sum
of squares within’ (SSW).
– The variability between groups is measured by the
‘sum of squares between’ (SSB).
Mean Square Within and Mean
Square Between
• The mean squares are measures of the average
variability and are calculated from the sum of
squares divided by the degrees of freedom.
• MSW = SSW/(N-j)
– MSW has N-j degrees of freedom where N= total
number of observations and j= # groups
• MSB = SSB / ( j-1)
– MSB = j-1 degrees of freedom where j= number of
groups
• F-statistic = MSB / MSW
ANOVA Assumptions
• The observations are from a random sample and they are
independent from each other
• The observations are normally distributed within each
group
• ANOVA is still appropriate if this assumption is not met when the
sample size in each group is at least 30
• The variances are approximately equal between groups
• If the ratio of the largest SD / smallest SD < 2, this assumption is
considered to be met.
• It is not required to have equal sample sizes in all
groups.
ANOVA example: mobility
The hypothetical data below represent mobility scores (higher score
indicates improved mobility) for 3 groups of patients:
Control group did not receive any therapy
Treatment group 1 received physical therapy,
Treatment group 2 received counseling and physical therapy.
Assume that the mobility scores are normally distributed.
Control Trt. 1 Trt. 2
35 38 47
38 43 53
42 45 42
34 52 45
28 40 46
39 46 37
ANOVA
• State the Hypotheses
– Null Hypothesis: µ control = µ trt 1 = µ trt 2
– Alternative Hypothesis: at least two of the means
(µ control , µ trt 1 , µ trt 2 ) are not equal
– ANOVA is always a two-sided test
– ANOVA will identify if at least two means are
significantly different but will not identify which two
(or more) means are different.
• Set the significance level
– α = 0.05.
Check for Equality of Variance
between groups
• Calculate the SD for each group
– Control: 4.86
– Trt 1: 4.94
– Trt 2: 5.33
• Calculate the ratio:
– largest SD / smallest SD = 5.33 / 4.86 = 1.1
• Since the ratio < 2, assume equality of
variance between groups.
Calculate Overall Mean and Group
Means
• Overall mean = average of all 18 observations =
41.7
• Group means = average of the observations in
each group
– Control mean = 36
– Treatment 1 mean = 44
– Treatment 2 mean = 45
Calculate the Within Sum of
Squares: SSW
• Square the difference between each observation and it’s
group mean and sum the 18 terms.
• The SSW for the control group (mean = 36)
(35 - 36)2 + (38 – 36)2 + (42 - 36)2 + (34 – 36)2 + (28 – 36)2 + (39 – 36)2 =
118
• The SSW for the treatment 1 group (mean = 44)
(38 - 44)2 + (43 – 44)2 + (45 - 44)2 + (52 – 44)2 + (40 – 44)2 + (46 – 44)2 =
122
• The SSW for the treatment 2 group (mean = 45)
(47 - 45)2 + (53 – 45)2 + (42 - 45)2 + (45 – 45)2 + (46 – 45)2 + (37 – 45)2 =
142
• SSW = sum of SSW for each group = 118 + 122 + 142 = 382
Calculate the Between Sum of
Squares: SSB
• The overall mean = 41.7
• The three group means are:
– Control: mean = 36
– Treatment 1: mean = 44
– Treatment 2: mean = 45
• For each group square the difference between the group mean and
the overall mean and multiply by the group sample size, then sum
these 3 terms for the SSB:

SSB = 6*(36 - 41.7)2 + 6*(44 – 41.7)2 + 6*(45 - 41.7)2 = 292


Calculate MSB and MSW
• The Mean Square Between (MSB) is the the
average variability between groups
– MSB = SSB /( j-1)
– SSB = 292, j = 3
– MSB = 292 / (3-1) = 146
• The Mean Square Within (MSW) is the average
variability within groups:
– MSW = SSW /(N-j)
– SSW = 382, N = 18, j= 3
– MSW = 382 / (18 – 3) = 25.47
Calculate the ANOVA F-statistic
• The ANOVA F-statistic = MSB/MSW
• The ANOVA F-statistic will be large when there is
more variability between the groups than within the
groups.
• If the variability between groups and within groups is
approximately equal the ANOVA F-statistic will be
small (close to 1.0)
• The Null hypothesis of equal means between groups is
rejected if the F-statistic is large enough.

• ANOVA F-statistic for example = 146/25.47 = 5.73


Sampling Distribution of
ANOVA F-statistic
The sampling distribution of the ANOVA F-statistic
is the F-distribution
– non-negative since all F-statistics are positive.
– indexed by two degrees of freedom
• Numerator df = number of groups minus 1 (j-1)
• Denominator df = total sample size minus number of groups
(N-j)
– The shape of the F-distribution varies depending on the
two degrees of freedom
• see page 167 in text for F distribution with 2, 70 df
Find the p-value of the ANOVA
F-statistic
• The p-value of the ANOVA F-statistic is the right tail
area greater than the F-statistic under the
F-distribution with (num. df, den. df)
• The F-statistic for the example data = 5.73
• The df for the F-distribution are (3-1) = 2 for the
numerator and (18-3) = 15 for the denominator.
• Use the FDIST function in EXCEL to find the p-value
FDIST(F-statistic value, num. df, den.df):
FDIST(5.73, 2, 15) = 0.014
Decision about the Hypotheses
• Since the p-value of 0.014 < the significance level of
0.05, the null hypothesis of equality between all three
group means is rejected.
• We can conclude that that AT LEAST two of the
means are significantly different.
• How many of the means are significantly different?
Which of the means are different?
• Post-hoc tests are done to answer these questions.
ANOVA: Data Analysis Tool in
EXCEL
• Select ‘Anova: Single Factor’
• Highlight all the observations
• Indicate if groups are in columns or rows
• Check the ‘Labels in First Row’ if you highlighted
labels
• Select the alpha level for the F-test (usually 0.05)
• Select ‘Output Range’ and indicate a cell for the output
table to be placed
• Select ‘OK’
• The ANOVA table and summary statistics will be
placed in the indicated location.
ANOVA table for Mobility Data
Example
SUMMARY
Groups Count Sum Average Variance
Column 1 6 216 36 23.6
Column 2 6 264 44 24.4
Column 3 6 270 45 28.4

ANOVA
Source of Variation SS df MS F P-value F crit
Between Groups 292 2 146 5.732984 0.0141425 3.682317
Within Groups 382 15 25.46667

Total 674 17

SSB MSW=SSW/df MSB = SSB/df


SSW
SST
ANOVA table
• Open the Excel Biostat_Intro3 file
• The data for the Mobility example are
provided on the ANOVA page
• Use the Data Analysis Tool to generate the
ANOVA table
ANOVA: post-hoc comparisons
• A significant ANOVA F-test is evidence that not
all means are equal but it does not identify which
means are significantly different.
• Methods used to find group differences after the
ANOVA null hypothesis has been rejected are
called post-hoc tests.
– Post-hoc is Latin for ‘after-this’
• Post-hoc comparisons should only be done when
the ANOVA F-test is significant.
Adjustments for Multiple
Comparisons
• When multiple comparisons are being done it is
customary to adjust the significance level of each
individual comparison so that the overall
experiment significance level remains at 0.05
• For an ANOVA with 3 groups, there are 3
combinations of t-tests.
• A conservative adjustment (Bonferroni
adjustment) is to divide 0.05/3 so that alpha for
each test = 0.017. Each comparison will be
significant if the p-value < 0.017
Post-hoc comparison: Control
and Treatment 1
• Results of the two-sample t-test to compare mean
mobility scores between the control group and
treatment 1 group:
– Control mean score = 36
– Treatment 1 mean score = 44
– P-value for two-sample t-test = 0.0179
• This is not significant at the adjusted α-level of 0.017 but
would be considered marginally significant since it is close to
0.017.
• Conclusion: After adjusting for multiple comparisons, the
control group mean mobility score is marginally
significantly less than the mean mobility score for the
group that received physical therapy (p = 0.0179).
Post-hoc comparison: Control
and Treatment 2
• A two-sample t-test is done to compare mean mobility
scores between the control group and treatment 2
group:
– Control mean score = 36
– Treatment 2 mean score = 45
– P-value for two-sample t-test = 0.012
• This is a significant difference at the adjusted α-level of 0.017
• Conclusion: After adjusting for multiple comparisons, the
control group mean mobility score is significantly less than the
mean mobility score for the group that received physical
therapy and counseling (p = 0.012).
Post-hoc comparison: Treatment 1
and Treatment 2
• A two-sample t-test is done to compare mean
mobility scores between the two treatment groups.
– Treatment 1 mean score = 44
– Treatment 2 mean score = 45
– P-value for two-sample t-test = 0.743
• There is not a significant difference between the two treatment
groups (at either the adjusted or un-adjusted level).
• Conclusion: There is no significant difference in mean
mobility score between the two treatment groups
(p = 0.743).
ANOVA summary
ANOVA was done to evaluate differences in mean mobility
score between three groups: a control group, a group that
received physical therapy only and a group that received
physical therapy and counseling. The significant ANOVA F-
test result indicated that at least two of the mean mobility
scores were significantly different.
Post-hoc t-tests with adjusted α−level = 0.017 (Bonferroni
adjustment for multiple comparisons) were done. Results of
the post-hoc comparisons indicated a significant difference
between the control group and the treatment group with both
physical therapy and counseling (p = 0.012), a marginally
significant difference between the control group and the
treatment group with physical therapy only (p = 0.0179) and
no significant difference between the two treatment groups.
Post-hoc Comparisons in Excel
• Open the Biostat_Intro3 Excel file.
• There is a worksheet to complete the three
post-hoc comparison t-tests: ANOVA post-
hoc
• Use the two-sample t-test Assuming equal
variance for each comparison
• Use the p-value for the two-tailed test.
Multiple Comparison Tests
• Bonferroni procedure
• Duncan Multiple range test
• Dunnett’s multiple comparison test
• Newman-Keuls test
• Scheffe’s test
• Tukey’s test
• Holm t-test
Other ANOVA Procedures
• One-way ANOVA is Analysis of Variance for one
factor
• More than one factor can be used for a two, three or
four-way ANOVA
• A continuous variable can be added to the model – this
is Analysis of Covariance (ANCOVA)
• Repeated Measures ANOVA can handle replicated
measurements on the same observation unit (subject)
Readings and Assignments
• Optional reading
– Chapter 7. Questions about Means in Three or
More Groups (ANOVA) pgs. 162-175
• Homework 3 due tomorrow

You might also like