当期总投资增加0.326143单位。
(2)模型的拟合优度为R?0.890687,修正可决系数为R?0.877022,可见模型拟合效果不错。 F检验:对模型进行显著性检验,F统计量对应的P值为0,因此在??0.05的显著性水平上我们拒绝原假设H:????0,说明回归方程显著,即变量“已发行股票的上期期末价值”和“上期资本”存量联合起来确实对“当期总投资”有显著影响。
F?F?2,16??3.63
22023?t检验:针对H:??0?j?1,2,3?进行显著性检验。给定显著性水平??0.05,查表知t?16??2.12。由回归结果,??、??对应的t统计量的绝对值均大于2.12,所以拒绝H:??0?j?2,3?;但??对应的t统计量的绝对值小于2.12,在0.05的显著性水平上不能拒绝H:??0的原假设。 (3)R?0.877022
(4)I在2003年的预测值为1254.848,置信度为95%的预测区间为(1030.292,1479.405) sf?105.9276
0j?2230j1012i
2.4 设一元线性模型为r (i=1,2,…..,n) 其回归方程为Y???????X,证明残差满足下式
23.1?i?Yi?Y??SXY(Xi?X)SXX
231如果把变量X,X分别对X进行一元线性回归,由两者残差定义的X,X关于X的偏相关系数r满足:
r?rr r?23123.123213123.1(1?r212)(1?r312)解答:
(1)对一元线性模型,由OLS可得
?X??Y????????X?X??Y?Y??SS??X?X?ii2iXYXX
所以,
???Y???i?Yi?Y??Xi??ii??SS?Yi??Y?XYX??XiXYSXX?SXX?S?Yi?Y?XY?Xi?X?SXX
(2)偏相关系数是指在剔除其他解释变量的影响后,一个解释变量对被解释变量的影响。不妨假设X,X对X进行一元线性回归得到的回归方程分别为: ????X?eX??,X??????X?e
则,e,e就分别表示X,X在剔除X影响后的值。
所以X,X关于X的偏相关系数就是指e,e的简单相关系数。 所以,
??e?e??e?e? r?231212113121212231231121i12i223.1??e1i?e1???e2i?e2?1222因为e?0,er21?1i1i?0??X??,?22i221i?X1??X2i?X2?1i??X?X1???X?X??X?X???X?X???X?X?1212i2,r31???X?X???X?X??X?X? ??X?X???X?X?1i11i13i3221i13i32??X,???21i?X1??X3i?X3?2
令X则??1i?X1?x1i,X2i?X2?x2i,X3i?X3?x3i
2?r21?x?x2222i,??122?r311i?x?x1223i
11i注意到X??????X,X??????X,所以e?x所以r???e?e??e?e???ee 1321i2i?2x1i,e2i?x3i???2x1i??1i12i21i2i23.1??e1i?e1???e2i?e2?22?e?e21i22i其中,
?ee??x1i2i?2x1i??x3i???2x1i????x2i???2?x1ix3i???2??2?x1i2x???2?x2ix1i??2x?3i2x?2i2x?2i2x?3i2i3i??x2ix3i?r31?r23?r23?r232?x23i21i?x2i1ix?r2123i?x2i21i?x21i1i3ix?r21?x22i1i231r1i?x23i21i?x21i21i?x?x2i?r31?x?x?x?x222i2i223i3i?r3121?r3121?xr?x?x?r?xr?x?x?rr?xr?x?x??r?rr?222122211i222i2i3i3i213123312123i?xr?x?x?r?x?x?rr?x?x?x?x231222i21313i3i2i222i2131r?x?x23i22i2同理可得:
?e?e21i2??1?r212??x2i2??1?r3123i2i ??x
2所以
r23.1??r23?r31r21??x2i2?x3i2?1?r??x?1?r??x22122i2312?r23?r31r213i?1?r??1?r?221231
2.7 2.7考虑下面两个模型: Ⅰ:Y????X???X???X?? Ⅱ:Y?X??????X????X????X??
i122illikkiiili122illikkii