中介效应后自变量的值变为负的如何解释

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

做中介效应，x为自变量，y为因变量，z为中介变量，y对x回归系数为正，z对x回归系数也为正，但y对x、z回归后x系数变为负的，这种情况可以出现吗，应如何解释？跪求大仙。。。。毕业论文

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：中介效应 自变量 回归系数 毕业论文 中介变量 自变量 如何 中介

I believe that there is a high correlation between the IV and the mediator in your research model. The results indicated that you did not have an effective mediation model since your IV and mediator are overlapping.

killsoftly 发表于 2016-10-5 12:52
I believe that there is a high correlation between the IV and the mediator in your research model. T ...

非常感谢解答，但是我还是有疑问，意思是我的中介变量选的不好对吗？还有自变量和中介变量的相关性应该在哪个范围合适呢？麻烦帮我看一下我选的变量间的相关性。另外，有啥好的建议调整吗，论文进展到结果分析了呢


             |    mj       mr       mt       mp     mc    enga     perfor
-------------+---------------------------------------------------------------
         mj |   1.0000
             |
             |
         mr |   0.5953   1.0000
             |   0.0000
             |
         mt |   0.5000   0.6013   1.0000
             |   0.0000   0.0000
             |
         mp |   0.5416   0.7072   0.6536   1.0000
             |   0.0000   0.0000   0.0000
             |
         mc |   0.4172   0.5402   0.5840   0.5676   1.0000
             |   0.0000   0.0000   0.0000   0.0000
             |
中介 enga |   0.4459   0.5782   0.5515   0.5939   0.6564   1.0000
             |   0.0000   0.0000   0.0000   0.0000   0.0000
             |
因perfor |   0.3718   0.4799   0.4680   0.5193   0.4616   0.7272   1.0000
             |   0.0000   0.0000   0.0000   0.0000   0.0000   0.0000

killsoftly 发表于 2016-10-5 12:52
I believe that there is a high correlation between the IV and the mediator in your research model. T ...

前五个都是自变量

树上栗子 发表于 2016-10-9 15:37
前五个都是自变量

Thanks for sharing your correlations

1. Having a look on the correlations among the IVs, the results actually reflect high correlations among your IVs.

2. The inter-correlations among your IVs are higher than that of the inter-correlations among IVs and the DV.

My suggestions:

1. Delete the construct which has the highest correlation - such as mr or mt !

2. or create a second-order construct that includes all IVs as the first-order constructs.

The second approach could possibly solve your problem since it reduces the threat of collinearity; however, I am not sure if you can justify the development of a second-order construct in your model based on your current theory or not. Thanks !  

killsoftly 发表于 2016-10-10 12:56
Thanks for sharing your correlations

1. Having a look on the correlations among the IVs, the re ...

非常感谢你的分析和建议，对我帮助很大，自变量间可能确实存在共线性的问题，限于问卷变量的设置，看来只能从自变量结构上调整了。另外，我看温忠麟有一篇文章《中介效应分析_方法和模型发展_温忠麟》在中介效应检验中提到了这种情况存在遮掩效应，不知道你对这个了解吗，求解。。

树上栗子 发表于 2016-10-11 08:08
非常感谢你的分析和建议，对我帮助很大，自变量间可能确实存在共线性的问题，限于问卷变量的设置， ...

You are welcome.

Multicollinearity is the usual suspect of suppression effects which means the high correlation among your independent variables leads to a wrong sign on one of the coefficients. Your case probably is caused by the suppression effects.

You could have a look on principal components regression or ridge regression. These two tools have a better ability to handle the problem of multicollinearity.

In addition, did you standardize your research variables prior to regression?

killsoftly 发表于 2016-10-14 13:35
You are welcome.

Multicollinearity is the usual suspect of suppression effects which means the ...

我一直对变量标准化处理存在疑问，看来又要请教你了，这次所用的变量没有标准化处理，如果进行标准化处理，是所有变量都处理吗？

树上栗子 发表于 2016-10-18 16:14
我一直对变量标准化处理存在疑问，看来又要请教你了，这次所用的变量没有标准化处理，如果进行标准化 ...

You suppose to standardize all variables prior to analysis. However, what is your question regarding standardized scores?

killsoftly 发表于 2016-10-20 12:19
You suppose to standardize all variables prior to analysis. However, what is your question regardi ...

You  are so nice.中期答辩刚结束，我会根据你的意见进行后续检验，但最近要先搁置一段时间了，投入找工作大军中.....，后续再有问题会继续请教大神哦，谢谢