In: Statistics and Probability
The drug ancrod was tested in a double-blinded clinical trial in which subjects who had strokes were randomly assigned to get either ancrod or a placebo. One outcome of interest in the study was whether or not a subject experienced intracranial hemorrhaging. In the ancrod group 13 of 248 experienced hemorrhaging, compared to 6 patients out of 252 in the placebo group.
a. Compute a 90% confidence interval for the difference of the proportions of patients experiencing hemorrhaging in the two groups.
• Point estimate
• Standard error
• Multiplier
• Confidence interval
• Check for large sample size
b. Conduct an appropriate test to determine whether the proportions of patients experiencing hemorrhaging in the two groups are different. Use a significance level of 0.01.
• Pooled sample proportion
• Test statistic
• Rejection region
• Decision
• Check for large sample size
first sample size, ancrod n1= 248
number of successes, sample 1 = x1=
13
proportion success of sample 1 , p̂1=
x1/n1= 0.0524
second sample size, placebo n2 = 252
number of successes, sample 2 = x2 =
6
proportion success of sample 1 , p̂ 2= x2/n2 =
0.024
Point estimate =difference in sample proportions, p̂1 - p̂2
= 0.0524 -
0.0238 = 0.0286
level of significance, α = 0.10
Z critical value = Z α/2 =
1.645 [excel function: =normsinv(α/2)
Std error , SE = SQRT(p̂1 * (1 - p̂1)/n1 + p̂2
* (1-p̂2)/n2) = 0.0171
multiplier =margin of error , E = Z*SE =
1.645 * 0.0171 =
0.0281
confidence interval is
lower limit = (p̂1 - p̂2) - E = 0.029
- 0.0281 = 0.0005
upper limit = (p̂1 - p̂2) + E = 0.029
+ 0.0281 = 0.0567
so, confidence interval is ( 0.0005
< p1 - p2 < 0.0567
)
b)
Ho: p1 - p2 = 0
Ha: p1 - p2 ╪ 0
sample #1 ----->
experimental
first sample size, n1=
248
number of successes, sample 1 = x1=
13
proportion success of sample 1 , p̂1=
x1/n1= 0.0524
sample #2 -----> standard
second sample size, n2 =
252
number of successes, sample 2 = x2 =
6
proportion success of sample 1 , p̂ 2= x2/n2 =
0.024
difference in sample proportions, p̂1 - p̂2 =
0.0524 - 0.0238 =
0.0286
pooled proportion , p =
(x1+x2)/(n1+n2)= 0.0380
std error ,SE = =SQRT(p*(1-p)*(1/n1+
1/n2)= 0.0171
Z-statistic = (p̂1 - p̂2)/SE = (
0.029 / 0.0171 ) =
1.6729
p-value =
0.0943 [excel formula
=2*NORMSDIST(z)]
decision : p-value>α,Don't reject null
hypothesis
THANKS
revert back for doubt
please upvote
Conclusion: There is not enough evidence to