1 Problem 1: STAT 3004: Solutions of Assignment 1
1 Problem 1: STAT 3004: Solutions of Assignment 1
Cheng Li
28 Sep 2020
1 Problem 1
1.1 (7.56)
We will use the paired t test
1.2 (7.57)
We wish to test the hypothesis:
H0 : µd = 0,
H1 : µd 6= 0,
where µd = mean of 4 year LVM - mean of Baseline LVM. So, we have the following equations,
d¯ = 18.9g,
sd = 26.4g,
Thus,
d¯ 18.9
t= = = 2.264,
√sd 26.4
√
10 10
The p-value = 2 ∗ P r(t9 > 2.264). From the distribution of t9 , we can get the p-value is less than 0.05.
Therefore, we can say that there is a significant increase in LVM over 4 years.
1.3 (7.58)
We first find a 95% CI for µd given by:
tn−1,.975 ∗ sd
d¯ ± √ = 18.9 ± t9,.975 ∗ 8.348 = 18.9 ± 2.262 ∗ 8.348 = (0, 37.8),
n
Thus, the CI for µd is (0, 37.8).
1.4 (7.59)
We use the sample size formula:
σ 2 (z1−α/2 + z1−β )2
n= ,
∆2
In this case, ∆ = 10g, σ is approximated by sd = 26.4g,z1−α/2 = z.975 = 1.96, z1−β = z.80 = 0.84. Thus,
n = 54.6 ≈ 55.
So, we need to study 55 subjects in the main study to achieve 80% power.
2 Problem 2
2.1 (7.12)
We have such a hypothesis:
H0 : p = p0 ,
H1 : p 6= p0 ,
where p0 = 0.005, p is the true incidence rate of MI in 2010 among 45-54-year-old men. Since np0 q0 =
5000 ∗ 0.005 ∗ 0.995 = 24.88 > 5, we can use the normal-theory method. So, we can get the test statistics
as following:
p̂ − p0
z=p
p0 q0 /n
15/5000 − 0.005
=p
0.005 ∗ 0.995/5000
= −2.005 < −1.96 = z0.025 ,
2.2 (7.13)
The p-value is:
p = 2 ∗ Θ(z)
= 2 ∗ (1 − Θ(1 − 2.005)
= 0.045.
2.3 (7.14)
From the problem, we can get a new hypothesis:
H0 : pdeath = p(0,death) ,
H1 : pdeath 6= p(0,death) ,
where p(0,death) = 0.25. Since nM I p(0,death) q(0,death) = 2.81 < 5, we must use the exact method to test
5
the hypothesis. Since p̂ = 15 > p(0,death) , the two-tailed p-value is obtained from
15
X 15
p=2∗ (0.25)k (0.75)(15−k)
k
k=5
4
X 15
= 2 ∗ (1 − (0.25)k (0.75)(15−k) )
k
k=0
= 0.627 > 0.05,
therefore, there is no significant change in the case-fatality rate between 2000 and 2010.
2
2.4 (7.15)
We can use the power formula in Equation 7.32 using a two-sided formulation whereby
√
|p0 − p1 | n
r
p 0 q0
P ower = Φ[ (zα/2 + √ ],
p 1 q1 p 0 q0
where p0 = 0.25, p1 = 0.2, α = 0.05, n = 50. So, we have
P ower = Φ(−1.238)
= 1 − Φ(1.238)
= 0.11,
thus, such a study would only have an 11% chance of detecting a significant difference.
2.5 (7.16)
We use the formula in Equation 7.33 using a two-sided formulation whereby
q
p0 q0 (z1−α/2 + z1−β pp10 qq10 )2
n= ,
(p1 − p0 )2
where β = 0.1. So, we have
0.1875[1.96 + 1.28(0.9238)]2
n=
0.0025
= 740.6 ≈ 741,
thus, we need to study 741 MI case to achieve 92% power.
3 Problem 3
3.1
From the problems, we can get the hypothesis
H0 : µ1 − µ2 = 0,
H1 : µ1 − µ2 6= 0,
and the estimators of sample variance s2 and sample µ̂ mean
(n1 − 1)s21 + (n2 − 1)s22
s2 = ,
n1 + n2 − 2
µ̂ = µˆ1 − µˆ2 ,
where µˆ1 = 6.56, s1 = 0.64, µˆ2 = 6.80 and s2 = 0.76.
Therefore, we can get the pivotal function
µ̂
t= q ,
s n11 + 1
n2
thus, t ≈ −1.313. Then, we can get the p-value is 0.1940 larger than 0.05. As a result, we think that there
is not a significant difference between the two groups.
3
3.2
Therefore, the CI is
r
1 1
µ̂ ± t( 63, 0.975) s2 ( + ),
n1 n2
4 Problem 4
4.1
We should use F-test to compare the standard deviation of diet record vitamin C intake between current
smokers vs. nonsmokers. The hpyothesis is
H0 : σ1 = σ2 ,
H1 : σ1 6= σ2 ,
and the F = ( ss12 )2 = 3.67, the p-value is 0.002 larger than 0.05, therefore we should reject the null
hpyothesis.
4.2
We need to use a two-sample t-test with unequal variance. So, we can get that
(x¯1 − x¯2
t= q 2 ,
s1 s22
306 + 17
= 5.07,
so, we take d0 = 23, p-value is less than 0.05. Thus, we reject null hypothesis.