分段回归与全样本回归分析结果比较

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

我想问：在分段回归中，两个关键解释变量在2000~2005年显著为负，在2005~2010不显著，而在全样本2000~2010年回归中显著为负，那我可以说这是由于在2000~2005年的数据所导致的吗？谢谢各位帮忙，如果可以的话，解析一下原因，不胜感激！

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏3 回帖

关键词：回归分析 不胜感激 解释变量 样本 回归分析 不胜感激

可以

lopez93 发表于 2014-2-28 18:35
可以

首先谢谢您！如果有解释说明的话，那就更感激！

diligentsai 发表于 2014-2-28 21:28
首先谢谢您！如果有解释说明的话，那就更感激！

我觉得嘛 你可以把被解释变量和那两个解释变量分别画图 看一下所谓的unconditional correlation是什么样的 直观地去观察你的数据 再去思考变化的原因是什么 这样可能在写文字阐述时会更加有理有据

if only using the two variables as independent variables, we can think the B (coefficient vector) based on all sample is a weighted average of B1 and B2 based on samples from 00-05 and 06-10, as an approximation, if B2 is not significant, then B depends on B1

jjjj6666 发表于 2014-2-28 23:28
if only using the two variables as independent variables, we can think the B (coefficient vector) ba ...

firstly, thank you ! if there are many variables including the above two critical explained ones and other controlled variables , but for B1,two critical explained variables are negatively significant ,for B2, they are nonsignificantly positive.Besides, there are not obviously different in other controled variables at two periods. therefore, can i conclude that according to my understanding of what you said, the negatively significance of  two critical explained variables in B depends on B1.As i am not good at english ,i explain  simple question into complicated one. i hope you can understand what i mean. thank you!

That's correct, the logic is the same, B based on overall sample can still be treated as weighted average of B1 and B2.  The other coefficients other than the ones based on x1 x2 depend on values from both B1 and B2, but for coefficients corresponding to x1 x2, the dominate ones are from samples based on 2000-2005.

jjjj6666 发表于 2014-3-1 12:16
That's correct, the logic is the same, B based on overall sample can still be treated as weighted av ...

thank you very much!

lopez93 发表于 2014-2-28 22:57
我觉得嘛 你可以把被解释变量和那两个解释变量分别画图 看一下所谓的unconditional correlation是什么样的 ...

有意思！