MPH 830 Problem Set # 3 Part 1 (ANOVA)

Type short answer when appropriate. For multiple choice, highlight or specify correct letter for
answer. Fill in the blanks where provided. Point distribution is indicated for each question. The
assignment is worth 90 points total.

1. (1 point) Describe the circumstances under which you should use ANOVA instead of t tests,
and explain why t tests are inappropriate in these circumstances.

Solution:

We use ANOVA when we want to compare 3 or more treatments. t-tests are limited to
comparing only 2 treatments. We cannot do multiple t tests because the probability of creating
a type 1 error increases due to t-tests utilizing test wise alpha levels, not experiment-wise alpha
levels.

2. (1 point) What value is expected, on average, for the F-ratio in ANOVA when the null
hypothesis is true? Explain why this value is expected.

Solution:

The expected value of the F-ratio in ANOVA when the null hypothesis is true is close to one to
be statistically significant.

If the null hypothesis of no group differences is true, then the expected value of the numerator
and denominator of the F-ratio will be equal.
3. (1 point) Describe the circumstances in which post tests are used and explain why these
tests are necessary.

Solution:

Post test are used to determine exactly which treatment conditions are significantly different.

It is also used when you reject the null hypothesis with more than 2 treatment conditions.

4. (20 points) The data below are from an independent-measures experiment comparing three
different treatment conditions.

Treatment 1 Treatment 2 Treatment 3

0 1 6

1 4 5 G = 36

0 1 8 ΣX 2 = 182

3 2 5

–––––––––––––––––––––––––––––––––––––

T=4 T=8 T = 24

SS = 6 SS = 6 SS = 6

a) Fill out the ANOVA table below:

Solution:

(𝑇12 + 𝑇22 + 𝑇32 ) 𝐺2 (42 +82 +242 ) 362

SSbetween = − = − = 56
𝑛 𝑁 4 12

𝐺2 362
SStotal = ∑𝑥 2 − = 182 − = 74
𝑁 12

SSwithin = SStotal – Ssbetween = 74 – 56 = 18

dfbetween = k -1 = 3 -1 = 2

dfwithin = N – k = 12 – 3 = 9

dftotal = N-1 = 12 - 1 = 11
 𝑆𝑆𝑏𝑒𝑡𝑤𝑒𝑒𝑛 56
MSb = = = 28
𝑑𝑓𝑏𝑒𝑡𝑤𝑒𝑒𝑛 2

𝑆𝑆𝑤𝑖𝑡ℎ𝑖𝑛 18
MSw = = =2
𝑑𝑓𝑤𝑖𝑡ℎ𝑖𝑛 9

𝑀𝑆𝐵 28
F= = = 14
𝑀𝑆𝑤 2

Source SS df MS F

Between 56 2 28 14

Within 18 9 2

Total 74 11

b) Use an analysis of variance with α = .05 to determine whether these data indicate any
significant mean differences among the treatments.

Fcritical = 4.2565

Decision = since F critical is less than F calculated, so reject H0.

5. (12 points) A researcher used an analysis of variance to compare three treatment conditions with a
separate sample of n = 8 participants in each treatment. The results of the analysis are shown in the
following summary. Fill in all missing values in the table. (Hint: Start with the df values.)

Source SS df MS
 Between Treatments ____ ____ F= 4

Within Treatments 63 ____ ____

Total ____ ____

Solution:

dfbetween = k -1 = 3 -1 = 2

dfwithin = N – k = 24 – 3 = 21

dftotal = N - 1 = 24 - 1 = 23

𝑆𝑆𝑤𝑖𝑡ℎ𝑖𝑛 63
MSw = = =3
𝑑𝑓𝑤𝑖𝑡ℎ𝑖𝑛 21

𝑀𝑆𝐵 = 𝐹 ∗ 𝑀𝑆𝑤 = 4 ∗ 3 = 12

𝑆𝑆𝑏𝑒𝑡𝑤𝑒𝑒𝑛 = MSb ∗ dfbetween = 12 ∗ 2 = 24

SStotal = SSwithin + Ssbetween = 63 + 24 = 87

6. (22 points) A psychologist would like to examine the effects of sugar consumption on the activity
level of preschool children. Three samples of children are selected with n = 5 in each sample. One
group gets no sugar, one group gets a small dose, and the third group gets a large dose. The
psychologist records the activity level for each child. The data from this experiment are presented
below.

No sugar Small dose Large dose

 5 3 6

3 3 8

1 7 5 G = 60

1 4 2 ΣX2 = 298

5 3 4

T = 15 T = 20 T = 25

SS = 16 SS = 12 SS = 20

a. Fill in the ANOVA table below:

Solution:

(𝑇12 + 𝑇22 + 𝑇32 ) 𝐺2 (152 +202 +252 ) 602

SSbetween = − = − = 10
𝑛 𝑁 5 15

𝐺2 602
SStotal = ∑𝑥 2 − = 298 − = 58
𝑁 15

SSwithin = SStotal – Ssbetween = 58 – 10 = 48

dfbetween = k -1 = 3 -1 = 2

dfwithin = N – k = 15 – 3 = 12

dftotal = N-1 = 15 - 1 = 14

𝑆𝑆𝑏𝑒𝑡𝑤𝑒𝑒𝑛 10
MSb = = =5
𝑑𝑓𝑏𝑒𝑡𝑤𝑒𝑒𝑛 2

𝑆𝑆𝑤𝑖𝑡ℎ𝑖𝑛 48
MSw = = =4
𝑑𝑓𝑤𝑖𝑡ℎ𝑖𝑛 12

𝑀𝑆𝐵 5
F= = = 1.25
𝑀𝑆𝑤 4

Source SS df MS F

Between 10 2 5 1.25

Within 48 12 4

Total 58 14
b. Do these data indicate any significant differences among the three groups? Test with α = .05.
Fcritical = 3.8853
Decision = since F critical is greater than F calculated, so we fail to reject H0.
c. Compute η2 , the percentage of variance accounted for by the treatment.

𝑆𝑆𝑏𝑒𝑡𝑤𝑒𝑒𝑛 10
η2 = = = 0.17
𝑆𝑆𝑡𝑜𝑡𝑎𝑙 58