In: Statistics and Probability
How can you reach a hypothesis-testing conclusion by examining a confidence interval for a two independent sample t-test? Is it possible? If so, how?
Yes, it is possible to reach a hypothesis-testing conclusion by examining a confidence interval for a two independent sample t-test.
Suppose we are interested to test if there is any significant difference in the means of two independent samples and our test statistic is the difference of sample means.
Let the significance level of the test is
, 0<
<1
and (1-)100%
confidence interval is (t1, t2). That means
if we take the sample 100 times and test the difference in means,
(1-
)100
times this interval will be able to contain the difference between
two population means and
100 times this interval is expected to fail to contain the true
difference in means.
Our Null hypothesis: true difference is 0 and alternative hypothesis: true difference is not 0.
If our test statistic value falls outside the confidence
interval, then we will reject the null hypothesis at 100%
level of significance. If it falls inside the confidence interval
then we fail to reject null hyothesis at 100
%
level of significance.