Consider the i.i.d. r.v.’s X_{1},\,\ldots,\,X_{20} with continuous d.f. F, which has unique x_{0.50},\mathcal{x}_{0.25},\mathrm{~and~}x_{0.75}, and let \,Y_{1},\,\ldots,\,Y_{20} be the corresponding order statistics. Then consider several confidence intervals for x_{0.50},x_{0.25},\mathrm{and}\;x_{0.75}, and calculate the respective coverage probabilities.
Using formula (11),
P(Y_{i}\leq x_{p}\leq Y_{j})=\sum\limits_{k=i}^{n}{\binom{n}{k}}p^{k}(1-p)^{n-k}-P(Y_{j}\leq x_{p})
~~=\sum\limits_{k=i}^{n}{\binom{n}{k}}p^{k}(1-p)^{n-k}-\sum\limits_{k=j}^{n}{\binom{n}{k}}p^{k}(1-p)^{n-k}
~~=\sum\limits_{k=i}^{j-1}{\binom{n}{k}}p^{k}(1-p)^{n-k}. (11)
we obtain the coverage probabilities listed in Table 15.1 for several confidence intervals for the median x_{0.50} and the first quartile x_{0.25}. For the calculation of coverage probabilities for confidence intervals for the third quartile x_{0.75}, we employ the following formula, which allows us to use the Binomial tables; namely,
\sum\limits_{k=i}^{j-1}{\binom{20}{k}}(0.75)^{k}(0.25)^{20-k}=\sum\limits_{r=20-j+1}^{20-i}{\binom{20}{r}}(0.25)^{r}(0.75)^{20-r}.
Table 15.1 | ||
Quantile | Confidence Interval | Coverage Probability |
x_{0.50} | (Y_{9},Y_{12}) | 0.3364 |
(Y_{8},Y_{13}) | 0.6167 | |
(Y_{7},Y_{14}) | 0.8107 | |
(Y_{6},Y_{15}) | 0.9216 | |
x_{0.25} | (Y_{6},\,Y_{6}) | 0.2024 |
(Y_{3},Y_{7}) | 0.5606 | |
(Y_{2},\,Y_{8}) | 0.8069 | |
(Y_{1},\,Y_{9}) | 0.9348 | |
x_{0.75} | (Y_{15},Y_{17}) | 0.2024 |
(Y_{14},Y_{18}) | 0.5606 | |
(Y_{13},Y_{19}) | 0.8069 | |
(Y_{12},Y_{20}) | 0.9348 |