0

Values of pressure of 10 patients before and after the administration of aspirin is given. Using sign test it is to be tested if aspirin was effective. There problem I am facing is that all the after values are evidently lower then before values. If before-after values are obtained then, the number of positive values is obviously 10. How do I conclude?

user97428
  • 1
  • Why would the number of positive values being 10 cause any more difficulty in figuring out what to do than there being say 9 positive values? Can you be more specific about where you have trouble? – Glen_b Dec 07 '15 at 23:46
  • Welcome to Cross Validated! Please add the `self-study` tag, read its [tag-wiki](http://stats.stackexchange.com/tags/self-study/info) and modify your question to follow the guidelines on asking such questions. In particular, you'll need to clearly identify what you've done to solve the problem yourself, and indicate the specific help you need at the point you struck difficulty. – Glen_b Dec 07 '15 at 23:47

1 Answers1

1

It sounds like you are facing a situation where you may end up concluding that there is a difference between the values of blood pressure before and after the implementation of the treatment.

Using a Wilcoxon signed-rank test on made up, but hopefully plausible:

      before    after
1   97.31477 82.81797
2   89.36883 75.00495
3   97.64900 73.26171
4   97.36215 77.55269
5   93.07321 77.50392
6   83.30025 76.94245
7   86.35716 80.26112
8   89.52640 74.54039
9   90.97116 81.17842
10 103.02327 72.81231

you get a "statistically significant" result:

wilcox.test(before, after, alternative = "two.sided", exact = T)

    Wilcoxon rank sum test

data:  before and after
W = 100, p-value = 1.083e-05
alternative hypothesis: true location shift is not equal to 0

Perhaps a secondary question is whether it would be sensible to run a paired t-test, since biological parameters like blood pressure are known to follow a normal distribution:

t.test(before, after, paired = T, alternative = "two.sided")

    Paired t-test

data:  before and after
t = 6.4866, df = 9, p-value = 0.0001132
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 10.16422 21.04983
sample estimates:
mean of the differences 
               15.60703

The small size of the sample in this case may not be prohibitive, and in this regard, this post on the t-test and the number of subjects in the sample is quite enlightening.

The advantage in this case wouldn't be the increased power of a parametric test, but rather the inclusion of some measure of the magnitude of the effect, such as the confidence interval of the difference: 95 percent confidence interval: 10.16422 - 21.04983.

Community
  • 1
Antoni Parellada
  • 23,430
  • 15
  • 100
  • 197
  • This is interesting, though the question relates to issues specifically with the sign test; you'd probably need to address the test in the question in some way. – Glen_b Dec 08 '15 at 06:13