题目
Problem
A machine fills bottles with mineral water.
The machine is checked every day to ensure that it is working correctly. On a particular
day a random sample of 100 bottles is taken. The volume of water, x millilitres, for each
bottle is measured and each measurement is coded using
y = x − 1000 y=x-1000 y = x − 1000
The results are summarised below
∑ y = 847 ∑ y 2 = 13 510.09 \sum y=847 \qquad \sum y^2=13\ 510.09 ∑ y = 847 ∑ y 2 = 13 510.09
(a) (i) Show that the value of the unbiased estimate of the mean of x is 1008.47
(ii) Calculate the unbiased estimate of the variance of x
(4)
The machine was initially set so that the volume of water in a bottle had a mean value
of 1010 millilitres.
Later, a test at the 5% significance level is used to determine whether or not the mean
volume of water in a bottle has changed. If it has changed then the machine is stopped
and reset.
(b) Write down suitable null and alternative hypotheses for a 2-tailed test.
(1)
(c) Find the critical region for X ‾ \overline X X in the above test.
(4)
(d) Using your answer to part (a) and your critical region found in part (c), comment on
whether or not the machine needs to be stopped and reset.
Give a reason for your answer.
(2)
(e) Explain why the use of σ 2 ≈ s 2 \sigma^2 \approx s^2 σ 2 ≈ s 2 is reasonable in this situation.
(1)
(Total for Question 1 is 12 marks)
题目中文翻译
一台机器正在灌装矿泉水瓶。
机器每天都会检查,以确保它正常工作。某天随机抽取了 100 个瓶子。
记录每个瓶中的水量 x x x 毫升,并使用
y = x − 1000 y=x-1000 y = x − 1000
进行编码。
结果汇总如下:
∑ y = 847 ∑ y 2 = 13 510.09 \sum y=847 \qquad \sum y^2=13\ 510.09 ∑ y = 847 ∑ y 2 = 13 510.09
(a) (i) 证明 x x x 的无偏均值估计为 1008.47。
(ii) 求 x x x 的无偏方差估计。
机器最初被设定为每瓶水的平均容量为 1010 毫升。
后来在 5% 显著性水平下进行检验,以判断每瓶水的平均容量是否已经改变。
如果改变了,机器就会停止并重新设定。
(b) 写出合适的零假设和备择假设,用于双侧检验。
(c) 求上述检验中 X ‾ \overline X X 的临界域。
(d) 利用 (a) 小题的答案和 (c) 小题得到的临界域,说明机器是否需要停止并重新设定。
给出理由。
(e) 解释为什么在这种情况下使用 σ 2 ≈ s 2 \sigma^2 \approx s^2 σ 2 ≈ s 2 是合理的。
解答
(a)(i)
解法一
思路
展开
先由 ∑ y \sum y ∑ y 求编码后数据的样本均值 y ‾ \overline y y 。编码关系为 x = y + 1000 x=y+1000 x = y + 1000 ,所以解码时样本均值也加上 1000。
答题过程
展开
The sample mean of the coded values is
y ‾ = ∑ y 100 = 847 100 = 8.47. \begin{align*}
\overline y
=&\,\frac{\sum y}{100}\\
=&\,\frac{847}{100}\\
=&\,8.47.
\end{align*} y = = = 100 ∑ y 100 847 8.47.
Since x = y + 1000 x=y+1000 x = y + 1000 ,
x ‾ = y ‾ + 1000 = 8.47 + 1000 = 1008.47 , \begin{align*}
\overline x
=&\,\overline y+1000\\
=&\,8.47+1000\\
=&\,1008.47,
\end{align*} x = = = y + 1000 8.47 + 1000 1008.47 ,
as required.
解法二
思路
展开
官方评分资料也接受先把编码后的总和还原为原数据总和。100 个观测值各自都要加回 1000,因此 ∑ x = ∑ y + 100 ( 1000 ) \sum x=\sum y+100(1000) ∑ x = ∑ y + 100 ( 1000 ) 。
答题过程
展开
Using x = y + 1000 x=y+1000 x = y + 1000 ,
∑ x = ∑ y + 100 ( 1000 ) = 847 + 100000 = 100847. \begin{align*}
\sum x
=&\,\sum y+100(1000)\\
=&\,847+100000\\
=&\,100847.
\end{align*} ∑ x = = = ∑ y + 100 ( 1000 ) 847 + 100000 100847.
Therefore,
x ‾ = 100847 100 = 1008.47 , \overline x=\frac{100847}{100}=1008.47, x = 100 100847 = 1008.47 ,
as required.
(a)(ii)
解法一
思路
展开
所有数据都加上同一个常数不会改变方差,所以 x x x 与 y y y 的样本方差相同。直接用给出的 ∑ y \sum y ∑ y 、∑ y 2 \sum y^2 ∑ y 2 和分母 n − 1 n-1 n − 1 计算无偏样本方差。
答题过程
展开
Since x = y + 1000 x=y+1000 x = y + 1000 , adding the same constant to every observation does not change the variance. Hence
s x 2 = s y 2 . s_x^2=s_y^2. s x 2 = s y 2 .
Therefore,
∑ y 2 − ( ∑ y ) 2 n = 13510.09 − 847 2 100 = 6336. \begin{align*}
\sum y^2-\frac{(\sum y)^2}{n}
=&\,13510.09-\frac{847^2}{100}\\
=&\,6336.
\end{align*} ∑ y 2 − n ( ∑ y ) 2 = = 13510.09 − 100 84 7 2 6336.
Hence
s x 2 = s y 2 = 6336 99 = 64 . s_x^2=s_y^2=\frac{6336}{99}=\boxed{64}. s x 2 = s y 2 = 99 6336 = 64 .
解法二
思路
展开
也可以把 ∑ x 2 \sum x^2 ∑ x 2 一并还原,再直接使用原始数据的无偏方差公式。由 x = y + 1000 x=y+1000 x = y + 1000 展开 x 2 x^2 x 2 ,对 100 个观测值求和即可。
答题过程
展开
Since
x 2 = y 2 + 2000 y + 1000000 , x^2=y^2+2000y+1000000, x 2 = y 2 + 2000 y + 1000000 ,
we have
∑ x 2 = ∑ y 2 + 2000 ∑ y + 100 ( 1000000 ) = 13510.09 + 2000 ( 847 ) + 100000000 = 101707510.09. \begin{align*}
\sum x^2
=&\,\sum y^2+2000\sum y\\
&\,\hspace{2pt}+100(1000000)\\
=&\,13510.09+2000(847)\\
&\,\hspace{2pt}+100000000\\
=&\,101707510.09.
\end{align*} ∑ x 2 = = = ∑ y 2 + 2000 ∑ y + 100 ( 1000000 ) 13510.09 + 2000 ( 847 ) + 100000000 101707510.09.
From part (a)(i), x ‾ = 1008.47 \overline x=1008.47 x = 1008.47 . Thus the corrected sum of squares is
∑ x 2 − n x ‾ 2 = 101707510.09 − 100 ( 1008.47 ) 2 = 6336. \begin{align*}
\sum x^2-n\overline x^{\,2}
=&\,101707510.09\\
&\,\hspace{2pt}-100(1008.47)^2\\
=&\,6336.
\end{align*} ∑ x 2 − n x 2 = = 101707510.09 − 100 ( 1008.47 ) 2 6336.
Hence
s x 2 = 6336 99 = 64 . s_x^2=\frac{6336}{99}=\boxed{64}. s x 2 = 99 6336 = 64 .
(b)
解法一
思路
展开
题目要判断平均灌装量是否“发生改变”,可能变大也可能变小,因此使用双尾检验,并以机器原先设定的总体均值 1010 ml 作为原假设。
答题过程
展开
Let μ \mu μ be the population mean volume of water in a bottle. The hypotheses are
H 0 : μ = 1010 , H 1 : μ ≠ 1010. \begin{aligned}
H_0:&\ \mu=1010,\\
H_1:&\ \mu\ne1010.
\end{aligned} H 0 : H 1 : μ = 1010 , μ = 1010.
(c)
解法一
思路
展开
由 (a) 得 s 2 = 64 s^2=64 s 2 = 64 ,所以用 s = 8 s=8 s = 8 估计总体标准差。样本量为 100,样本均值的标准误为 8 / 100 = 0.8 8/\sqrt{100}=0.8 8/ 100 = 0.8 。双尾 5 % 5\% 5% 检验使用临界值 ± 1.96 \pm1.96 ± 1.96 ,把标准化不等式还原为 X ‾ \overline X X 的范围。
答题过程
展开
Under H 0 H_0 H 0 , using s = 8 s=8 s = 8 as an estimate of σ \sigma σ ,
X ‾ ≈ N ( 1010 , 8 2 100 ) . \overline X\approx
\mathrm N\left(1010,\frac{8^2}{100}\right). X ≈ N ( 1010 , 100 8 2 ) .
For a two-tailed test at the 5 % 5\% 5% significance level, the critical z z z -values are ± 1.96 \pm1.96 ± 1.96 . The lower boundary is
X ‾ = 1010 − 1.96 ( 8 100 ) = 1008.432 , \begin{align*}
\overline X
=&\,1010-1.96\left(\frac8{\sqrt{100}}\right)\\
=&\,1008.432,
\end{align*} X = = 1010 − 1.96 ( 100 8 ) 1008.432 ,
and the upper boundary is
X ‾ = 1010 + 1.96 ( 8 100 ) = 1011.568. \begin{align*}
\overline X
=&\,1010+1.96\left(\frac8{\sqrt{100}}\right)\\
=&\,1011.568.
\end{align*} X = = 1010 + 1.96 ( 100 8 ) 1011.568.
Therefore, the critical region is
X ‾ ⩽ 1008.432 or X ‾ ⩾ 1011.568 . \boxed{
\overline X\leqslant1008.432
\quad\text{or}\quad
\overline X\geqslant1011.568
}. X ⩽ 1008.432 or X ⩾ 1011.568 .
(d)
解法一
思路
展开
将 (a) 得到的样本均值 1008.47 1008.47 1008.47 与 (c) 的两个临界边界比较。它略高于下侧边界,因此没有落入临界域。
答题过程
展开
The observed sample mean is
x ‾ = 1008.47. \overline x=1008.47. x = 1008.47.
Since
1008.432 < 1008.47 < 1011.568 , 1008.432<1008.47<1011.568, 1008.432 < 1008.47 < 1011.568 ,
the sample mean does not lie in the critical region. Therefore, H 0 H_0 H 0 is not rejected and the machine does not need to be stopped and reset.
(e)
解法一
思路
展开
样本量 100 较大,因此样本方差通常能较稳定地估计总体方差,以 s 2 s^2 s 2 代替未知的 σ 2 \sigma^2 σ 2 是合理的。
答题过程
展开
It is reasonable to use σ 2 ≈ s 2 \sigma^2\approx s^2 σ 2 ≈ s 2 because the sample size is large.