题目
Problem
A large company rents shops in different parts of the country. A random sample of 10 shops was taken and the floor area, x x x in 10 m 2 10m^2 10 m 2 , and the annual rent, y y y in thousands of dollars, were recorded.
The data are summarised by the following statistics
∑ x = 900 ∑ x 2 = 84818 ∑ y = 183 ∑ y 2 = 3434 \sum x=900\qquad \sum x^2=84818\qquad \sum y=183\qquad \sum y^2=3434 ∑ x = 900 ∑ x 2 = 84818 ∑ y = 183 ∑ y 2 = 3434
and the regression line of y y y on x x x has equation y = 6.066 + 0.136 x y=6.066+0.136x y = 6.066 + 0.136 x
(a) Use the regression line to estimate the annual rent in dollars for a shop with a floor area of 800 m 2 800m^2 800 m 2
(2)
(b) Find S y y S_{yy} S y y and S x x S_{xx} S xx
(3)
(c) Find the product moment correlation coefficient between y y y and x x x .
(4)
An 11th shop is added to the sample. The floor area is 900 m 2 900m^2 900 m 2 and the annual rent is 15 000 dollars.
(d) Use the formula S x y = ∑ ( x − x ˉ ) ( y − y ˉ ) S_{xy}=\sum(x-\bar x)(y-\bar y) S x y = ∑ ( x − x ˉ ) ( y − y ˉ ) to show that the value of S x y S_{xy} S x y for the 11 shops will be the same as it was for the original 10 shops.
(2)
(e) Find the new equation of the regression line of y y y on x x x for the 11 shops.
(3)
The company is considering renting a larger shop with area of 3000 m 2 3000m^2 3000 m 2
(f) Comment on the suitability of using the new regression line to estimate the annual rent. Give a reason for your answer.
(1)
解答
(a)
解法一
思路
展开
x x x 的单位是 10 m 2 10m^2 10 m 2 ,所以 800 m 2 800m^2 800 m 2 对应 x = 80 x=80 x = 80 。代入后 y y y 的单位是 thousands of dollars。
答题过程
展开
For a floor area of 800 m 2 800m^2 800 m 2 ,
x = 80. \begin{align*}
x=80.
\end{align*} x = 80.
Using the regression line,
y = 6.066 + 0.136 ( 80 ) = 16.946. \begin{align*}
y=6.066+0.136(80)=16.946.
\end{align*} y = 6.066 + 0.136 ( 80 ) = 16.946.
Since y y y is in thousands of dollars, the estimated annual rent is
$ 16946. \begin{align*}
\$16946.
\end{align*} $16946.
(b)
解法一
思路
展开
分别使用 S y y = ∑ y 2 − ( ∑ y ) 2 n S_{yy}=\sum y^2-\frac{(\sum y)^2}{n} S y y = ∑ y 2 − n ( ∑ y ) 2 和 S x x = ∑ x 2 − ( ∑ x ) 2 n S_{xx}=\sum x^2-\frac{(\sum x)^2}{n} S xx = ∑ x 2 − n ( ∑ x ) 2 。
答题过程
展开
S y y = ∑ y 2 − ( ∑ y ) 2 n = 3434 − 183 2 10 = 85.1. \begin{aligned}
S_{yy}
=&\,\sum y^2-\frac{(\sum y)^2}{n}\\
=&\,3434-\frac{183^2}{10}\\
=&\,85.1.
\end{aligned} S y y = = = ∑ y 2 − n ( ∑ y ) 2 3434 − 10 18 3 2 85.1.
Also,
S x x = ∑ x 2 − ( ∑ x ) 2 n = 84818 − 900 2 10 = 3818. \begin{aligned}
S_{xx}
=&\,\sum x^2-\frac{(\sum x)^2}{n}\\
=&\,84818-\frac{900^2}{10}\\
=&\,3818.
\end{aligned} S xx = = = ∑ x 2 − n ( ∑ x ) 2 84818 − 10 90 0 2 3818.
(c)
解法一
思路
展开
回归线 y y y on x x x 的斜率 b = S x y S x x b=\frac{S_{xy}}{S_{xx}} b = S xx S x y ,先求 S x y S_{xy} S x y ,再求相关系数。
答题过程
展开
For the regression line of y y y on x x x ,
b = S x y S x x . \begin{align*}
b=\frac{S_{xy}}{S_{xx}}.
\end{align*} b = S xx S x y .
Hence
S x y = 0.136 ( 3818 ) = 519.248. \begin{align*}
S_{xy}=0.136(3818)=519.248.
\end{align*} S x y = 0.136 ( 3818 ) = 519.248.
Now
r = S x y S x x S y y = 519.248 3818 ( 85.1 ) = 0.910944 … . \begin{aligned}
r
=&\,\frac{S_{xy}}{\sqrt{S_{xx}S_{yy}}}\\
=&\,\frac{519.248}{\sqrt{3818(85.1)}}\\
=&\,0.910944\ldots.
\end{aligned} r = = = S xx S y y S x y 3818 ( 85.1 ) 519.248 0.910944 … .
Therefore
r = 0.911 to 3 significant figures . \begin{align*}
r=0.911\quad\text{to 3 significant figures}.
\end{align*} r = 0.911 to 3 significant figures .
(d)
解法一
思路
展开
原本 x ˉ = 900 10 = 90 \bar x=\frac{900}{10}=90 x ˉ = 10 900 = 90 。新 shop 的 x x x 也是 90,所以加入后新的 x ˉ \bar x x ˉ 仍是 90。新增项中的 x − x ˉ = 0 x-\bar x=0 x − x ˉ = 0 ,对 S x y S_{xy} S x y 没有贡献。
答题过程
展开
For the original 10 shops,
x ˉ = 900 10 = 90. \begin{align*}
\bar x=\frac{900}{10}=90.
\end{align*} x ˉ = 10 900 = 90.
For the 11 shops,
x ˉ = 900 + 90 11 = 90. \begin{align*}
\bar x=\frac{900+90}{11}=90.
\end{align*} x ˉ = 11 900 + 90 = 90.
The new shop has x = 90 x=90 x = 90 , so its contribution to
S x y = ∑ ( x − x ˉ ) ( y − y ˉ ) \begin{align*}
S_{xy}=\sum(x-\bar x)(y-\bar y)
\end{align*} S x y = ∑ ( x − x ˉ ) ( y − y ˉ )
is
( 90 − 90 ) ( 15 − y ˉ ) = 0. \begin{align*}
(90-90)(15-\bar y)=0.
\end{align*} ( 90 − 90 ) ( 15 − y ˉ ) = 0.
Therefore S x y S_{xy} S x y is unchanged.
解法二
思路
展开
计算简化公式法。
除了使用 S x y S_{xy} S x y 的定义式进行证明外,我们也可以使用其计算简化公式:
S x y = ∑ x y − ∑ x ∑ y n \begin{align*}
S_{xy} = \sum xy - \frac{\sum x \sum y}{n}
\end{align*} S x y = ∑ x y − n ∑ x ∑ y
我们可以对比原 10 个商店与加入第 11 个商店后新系统的这三部分变化:
对于 10 个商店:n = 10 n=10 n = 10 ,∑ x = 900 \sum x=900 ∑ x = 900 ,∑ y = 183 \sum y=183 ∑ y = 183 ,因此 S x y = ∑ x y 10 − 900 × 183 10 = ∑ x y 10 − 16470 S_{xy} = \sum xy_{10} - \frac{900 \times 183}{10} = \sum xy_{10} - 16470 S x y = ∑ x y 10 − 10 900 × 183 = ∑ x y 10 − 16470 。
对于 11 个商店:新加入的数据为 x 11 = 90 x_{11}=90 x 11 = 90 ,y 11 = 15 y_{11}=15 y 11 = 15 (注意年租金 \ 15000对应 对应 对应 y=15)。新样本数 )。
新样本数 )。新样本数 n’=11, , , \sum x’ = 900+90 = 990, , , \sum y’ = 183+15 = 198,新乘积和 ,新乘积和 ,新乘积和 \sum xy_{11} = \sum xy_{10} + 90 \times 15 = \sum xy_{10} + 1350。将上述数值代入 11 个商店的 。
将上述数值代入 11 个商店的 。将上述数值代入 11 个商店的 S_{xy}简化公式中展开计算,即可证明新旧 简化公式中展开计算,即可证明新旧 简化公式中展开计算,即可证明新旧 S_{xy}$ 恒等。许多同学更习惯使用计算简化公式,此方法提供了一个完全可行且严密的代数证明视角。
答题过程
展开
Use the computational formula for S x y S_{xy} S x y :
S x y = ∑ x y − ∑ x ∑ y n . \begin{align*}
S_{xy} = \sum xy - \frac{\sum x \sum y}{n}.
\end{align*} S x y = ∑ x y − n ∑ x ∑ y .
For the original 10 10 10 shops:
n = 10 n = 10 n = 10
∑ x = 900 \sum x = 900 ∑ x = 900
∑ y = 183 \sum y = 183 ∑ y = 183
Let ∑ x y 10 \sum xy_{10} ∑ x y 10 be the sum of products for the original 10 10 10 shops. The original S x y S_{xy} S x y is:
S x y , old = ∑ x y 10 − 900 × 183 10 = ∑ x y 10 − 16470. \begin{align*}
S_{xy,\text{old}} =&\,\, \sum xy_{10} - \frac{900 \times 183}{10}\\[3mm]
=&\,\, \sum xy_{10} - 16470.
\end{align*} S x y , old = = ∑ x y 10 − 10 900 × 183 ∑ x y 10 − 16470.
For the 11 11 11 shops, the new shop has x 11 = 90 x_{11} = 90 x 11 = 90 and y 11 = 15 y_{11} = 15 y 11 = 15 (since rent is in thousands of dollars).
The new summary statistics are:
n ′ = 11 n' = 11 n ′ = 11
∑ x ′ = 900 + 90 = 990 \sum x' = 900 + 90 = 990 ∑ x ′ = 900 + 90 = 990
∑ y ′ = 183 + 15 = 198 \sum y' = 183 + 15 = 198 ∑ y ′ = 183 + 15 = 198
∑ x y 11 = ∑ x y 10 + 90 × 15 = ∑ x y 10 + 1350 \sum xy_{11} = \sum xy_{10} + 90 \times 15 = \sum xy_{10} + 1350 ∑ x y 11 = ∑ x y 10 + 90 × 15 = ∑ x y 10 + 1350
The new S x y S_{xy} S x y is:
S x y , new = ∑ x y 11 − ∑ x ′ ∑ y ′ n ′ = ( ∑ x y 10 + 1350 ) − 990 × 198 11 = ∑ x y 10 + 1350 − 90 × 198 = ∑ x y 10 + 1350 − 17820 = ∑ x y 10 − 16470. \begin{align*}
S_{xy,\text{new}} =&\,\, \sum xy_{11} - \frac{\sum x' \sum y'}{n'}\\[3mm]
=&\,\, \left(\sum xy_{10} + 1350\right) - \frac{990 \times 198}{11}\\[3mm]
=&\,\, \sum xy_{10} + 1350 - 90 \times 198\\[3mm]
=&\,\, \sum xy_{10} + 1350 - 17820\\[3mm]
=&\,\, \sum xy_{10} - 16470.
\end{align*} S x y , new = = = = = ∑ x y 11 − n ′ ∑ x ′ ∑ y ′ ( ∑ x y 10 + 1350 ) − 11 990 × 198 ∑ x y 10 + 1350 − 90 × 198 ∑ x y 10 + 1350 − 17820 ∑ x y 10 − 16470.
Since S x y , new = S x y , old S_{xy,\text{new}} = S_{xy,\text{old}} S x y , new = S x y , old , the value of S x y S_{xy} S x y remains unchanged.
(e)
解法一
思路
展开
由于 S x y S_{xy} S x y 不变,且新 shop 的 x x x 等于平均 x x x ,S x x S_{xx} S xx 也不变,所以斜率仍是 0.136。新的 y ˉ \bar y y ˉ 用 11 个 shops 计算。
答题过程
展开
The gradient remains
b = 0.136. \begin{align*}
b=0.136.
\end{align*} b = 0.136.
The new mean of y y y is
y ˉ = 183 + 15 11 = 18. \begin{align*}
\bar y=\frac{183+15}{11}=18.
\end{align*} y ˉ = 11 183 + 15 = 18.
The new mean of x x x is still
x ˉ = 90. \begin{align*}
\bar x=90.
\end{align*} x ˉ = 90.
The new mean of y y y is 0.3 0.3 0.3 less than the original mean of y y y , while x ˉ \bar x x ˉ and the gradient are unchanged.
So the intercept is also 0.3 0.3 0.3 less than before:
a = 6.066 − 0.3 = 5.766. \begin{align*}
a=6.066-0.3=5.766.
\end{align*} a = 6.066 − 0.3 = 5.766.
Therefore the new regression line is
y = 5.766 + 0.136 x . \begin{align*}
y=5.766+0.136x.
\end{align*} y = 5.766 + 0.136 x .
(f)
解法一
思路
展开
3000 m 2 3000m^2 3000 m 2 对应 x = 300 x=300 x = 300 ,远大于样本平均 x = 90 x=90 x = 90 ,属于外推,因此不适合。
答题过程
展开
For a floor area of 3000 m 2 3000m^2 3000 m 2 ,
x = 300. \begin{align*}
x=300.
\end{align*} x = 300.
This is far outside the range of the data, compared with the mean value x ˉ = 90 \bar x=90 x ˉ = 90 .
Therefore using the regression line is not suitable, because it would involve extrapolation.