How to get marginal effects on joint prob and conduct fixed effects using gsem?

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

Dear all,


I have two questions:


1) I am trying to estimate the marginal effects using gsem, with probit as the option, do you know which command shall I use?


2) I also want to control for year and country fixed effects, can we directly put in the country or year dummies? Is there any problem in doing so?


My detailed problem is as follows:


I want to know whether international media shaming has effects on improving political rights, and whether government increases terror to offset the improvement in political rights, so I specified my models below:

1. global X = "l.cat_rat l.ccpr_rat l.democratic l.C_lgdppc l.C_lpop l.Civilwar2 l.War"
   
2. gsem (PIRI_2_increase <- l.Media_ai_tot l(1/3).PIRI_2 \$X  i.year i.idcode, family(binomial) link(probit))(polrt_decline <- l(2).Media_ai_tot l.polrt \$X  i.year i.idcode, family(binomial) link(probit)) (PIRI_2_increase <- l.polrt l(1/3).PIRI_2 \$X  i.year i.idcode, family(binomial) link(probit)) (Media_ai_tot <- l.polrt i.year i.idcode, regress) (polrt_decline <- l(1/3).PIRI_2 l.polrt \$X  i.year i.idcode, family(binomial) link(probit)), nocaps

复制代码

PIRI_2_increase is an indicator for whether terror has increased, Media_ai_tot is the level of international media shaming, polrt_decline is an indicator for whether political rights has improved (higher scores of PIRI and polirt mean worse cases). X are other covariates.


The first equation is to estimate the effects of media shaming on terror (probit model), the second is to estimate the effects of media shaming on political rights (probit model), the third is to estimate the effects of political rights on terror (probit model), the fourth is to estimate the effects of previous political rights on the amount of shaming (OLS), and the fifth is to estimate the effects of terror on political rights (probit).


I am particularly interested in knowing the marginal effects of media shaming in equation 1 and 2. I have figured out how to get the marginal effects for single equations.


For example, for equation 1:

1. estpost margins, dydx(l.Media_ai_tot) atmeans predict(pr outcome(PIRI_2_increase))

复制代码

I don't know how to get the joint probability of equation 1 and equation 2, that is, I want to know the marginal effects of shaming (l.Media_ai_tot) on the joint probability of terror increase (PIRI_2_increase=1 in equation 1) and political rights improvement (polrt_decline=1 in equation 2).


I would appreciate any help you could offer.

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0 回帖

关键词：fixed effect Marginal effects conduct Effect political detailed directly conduct command

From Stata's help document, it seems they do not have an option of predicting the joint probability. http://www.stata.com/manuals13/sempredictaftergsem.pdf

Just found out that we can use a bivariate probit model to get the joint probability, it works well, but the only problem is that we miss the information of the other 3 equations:

1. biprobit (f.PIRI_2_increase <- l.Media_ai_tot l(1/3).PIRI_2 \$X i.year) (f.polrt_decline <- l.Media_ai_tot l.polrt \$X i.year) if year>=1984&year<=2001, vce(robust)
   
2. 
   
3. estpost margins, dydx(l.Media_ai_tot) atmeans predict(p11)

复制代码