of the hypothesis 0 0 : H µ µ X Y − =∆ against the alternative hypothesis
1 0 : (, ) H µ µ X Y − ≠ ∆ . By means of Monte Carlo simulation, they found that the
test statistic T3 has better power than the test statistic T2 . However, they did not study
the confidence interval for θ using the pivotal test statistic T3 . Although the
correspondence between hypothesis testing, 0 0 : H µ µ X Y − =∆ against the
alternative hypothesis 1 0 : (, ) H µ µ X Y − ≠ ∆ , and the confidence interval
estimation had been well documented, the coverage probability and the expected length
of the confidence interval for θ , when one variance unknown, have not been
investigated.
As a result, it is of interest to construct the confidence interval for θ when one
variance is unknown. The proposed confidence interval is constructed using the pivotal
quantity T3 . Maity and Sherman [12] pointed out that the test statistic T3 has an
approximate t-distribution with ν 1 degrees of freedom where