凡是搞计量经济的，都关注这个号了
邮箱：econometrics666@sina.cn
所有计量经济圈方法论丛的code程序, 宏微观数据库和各种软件都放在社群里.欢迎到计量经济圈社群交流访问
.
下面是一些交乘项的相关文章，各位按需查看留存即可。
0.实证研究中交叉项的使用和解读策略指南案例
1.交互项与分组回归的区别是什么? 异质性分析
2.计量回归中的交互项到底什么鬼? 捎一本书给你
3.计量经济学中"交互项"相关的5个问题和回应
4.面板数据中去中心化的交互项回归什么情况
5.内生变量的交互项如何寻工具变量, 交互项共线咋办
6.省份/行业固定效应与年份固定效应的交乘项固定效应
7.U型和倒U型关系及其调节效应, 把非线性关系推进一点
8.中介和调节效应自助法检验，针对非正态截面数据
9.控制、调节和中介变量，系说
10.具有调节变量的中介效应程序和数据, 独家解读相关结果
11.具有调节变量的中介效应分析, moderated mediation
12.中介和调节效应操作指南, 经典书籍和PPT珍藏版
今天，咱们圈子引荐一篇关于交互项运用于Logit和Probit模型里的论文。简明地说，作者发现，使用margins去求Logit和Probit模型里交互项的边际效应的做法是不正确的，因此作者特别设计了一个新的程序“inteff”去求二值模型里交互项的边际效应，还能自动将交互图保存下来。
使用该程序也比较简单，只需要把logit后面那些变量全部放到inteff就好。不过，值得注意的是，对变量放置的顺序有些讲究，这个看看程序里的说明就知道了。
*Logit模型下交互项的边际效应
logit $y age educ ageeduc male ins_pub ins_uni $x, nolog cluster(pid)inteff $y age educ ageeduc male ins_pub ins_uni $x,savedata(d:\data\logit_inteff,replace)savegraph1(d:\data\figure1, replace)savegraph2(d:\data\figure2, replace)
*Probit模型下交互项的边际效应
probit $y male ins_uni male_uni age educ ins_pub $x, nolog cluster(pid)inteff $y male ins_uni male_uni age educ ins_pub $x,savedata(d:\data\probit_inteff, replace)savegraph1(d:\data\figure3, replace)savegraph2(d:\data\figure4, replace)
作者在Economics Letters表达了，他对传统通过margins求logit和probit模型里交互项的边际效应的不满，并且提出了新方法去求二值模型里交互项的边际效应。
作者在上面这篇文章发表一年后，就在Stata Journal上发表了这个新程序“inteff”。建议各位学者，在以后求Logit或Probit模型里交互项的边际效应时候，使用inteff而非margins程序。
Applied researchers often estimate interaction terms to infer how the effect of one independent variable on the dependent variable depends on the magnitude of another independent variable. For example, is the effect of car weight on gas mileage the same for both domestic and foreign cars? To answer this question, we can run a regression to predict gas mileage as a function of weight, a dummy variable for foreign, and the interaction between the two. If the coefficient on the interaction term is statistically significant, there is a difference between domestic and foreign cars in how additional weight affects mileage.
Interaction terms are also used extensively in nonlinear models, such as logit and probit models. Unfortunately, the intuition from linear regression models does not extend to nonlinear models. The marginal effect of a change in both interacted variables is not equal to the marginal effect of changing just the interaction term. More surprisingly, the sign may be different for different observations. The statistical significance cannot be determined from the z-statistic reported in the regression output. The odds-ratio interpretation of logit coefficients cannot be used for interaction terms.
Despite the common use of interaction terms, most applied researchers misinterpret the coefficient of the interaction term in nonlinear models. A review of 13 economics journals listed on JSTOR (www.jstor.org) found 72 articles published between 1980 and 2000 that used interaction terms in nonlinear models (Ai and Norton 2003). None of the studies interpreted the coefficient on the interaction term correctly. A recent article by DeLeire (2000) is a welcome exception.
The Stata command inteff computes the correct marginal effect of a change in two interacted variables for a logit or probit model. It also computes the correct standard errors. The inteff command will work if the interacted variables are both continuous variables, if both are dummy variables, or if there is one of each. In addition, it will graph the interaction effect and save the results to allow further investigation.
长按以上二维码可以查看全文
下面这些短链接文章属于合集，可以收藏起来阅读，不然以后都找不到了。
