有谁懂面板数据联立方程

Hi Everybody,

I am estimating a system of simultaneous equation using Proc Model and Proc
Syslin. I am trying to estimate my model with 2SLS and 3SLS. I am using SAS
9.1.3. For similar specification in both procedures I should get matching
results. However, I am not getting same results through both proceducures.
Could some one suggest what is going wrong? Is there anything wrong in
specification of model or am I not writing it properly in SAS.

proc model data=work.temp;
endogenous ordrs dllrs wslo;
parameters a a1 a2 a3 a4 a5 a6 b b1 b2 b3 b4 b5 b6 c c1 c2 c3 c4 c5 c6;

ordrs = a + a1*mp1 + a2*mp2 + a3*dllrs + a4*wslo + a5*totmld + a6*avmbmld;

dllrs = b + b1*mp1 + b2*mp2 + b3*ordrs + b4*wslo + b5*lhpa + b6*newhigh;

wslo = c + c1*mp1 + c2*mp2 + c3*ordrs + c4*dllrs + c5*loamt + c6*mllod;

instruments totmld avmbmld lhpa newhigh loamt mllod;

fit ordrs dllrs wslo / OLS 2sls 3sls;
run;

proc syslin data=work.temp ols 2sls 3sls;
endogenous ordrs dllrs wslo;
instruments totmld avmbmld lhpa newhigh loamt mllod;

model ordrs = mp1 mp2 dllrs wslo totmld avmbmld;

model dllrs = mp1 mp2 ordrs wslo lhpa newhigh;

model wslo = mp1 mp2 ordrs dllrs loamt mllod;

run;

以下是引用Trevor在2006-4-25 3:37:00的发言：

Hi Everybody,

I am estimating a system of simultaneous equation using Proc Model and Proc
Syslin. I am trying to estimate my model with 2SLS and 3SLS. I am using SAS
9.1.3. For similar specification in both procedures I should get matching
results. However, I am not getting same results through both proceducures.
Could some one suggest what is going wrong? Is there anything wrong in
specification of model or am I not writing it properly in SAS.

proc model data=work.temp;
endogenous ordrs dllrs wslo;
parameters a a1 a2 a3 a4 a5 a6 b b1 b2 b3 b4 b5 b6 c c1 c2 c3 c4 c5 c6;

ordrs = a + a1*mp1 + a2*mp2 + a3*dllrs + a4*wslo + a5*totmld + a6*avmbmld;

dllrs = b + b1*mp1 + b2*mp2 + b3*ordrs + b4*wslo + b5*lhpa + b6*newhigh;

wslo = c + c1*mp1 + c2*mp2 + c3*ordrs + c4*dllrs + c5*loamt + c6*mllod;

instruments totmld avmbmld lhpa newhigh loamt mllod;

fit ordrs dllrs wslo / OLS 2sls 3sls;
run;

proc syslin data=work.temp ols 2sls 3sls;
endogenous ordrs dllrs wslo;
instruments totmld avmbmld lhpa newhigh loamt mllod;

model ordrs = mp1 mp2 dllrs wslo totmld avmbmld;

model dllrs = mp1 mp2 ordrs wslo lhpa newhigh;

model wslo = mp1 mp2 ordrs dllrs loamt mllod;

run;

I don't see anything egregiously wrong with your code - assuming you have specified it correctly. And assuming this is not actually your code, but you have specified N2SLS and N3SLS instead of 2SLS and 3SLS in the PROC MODEL code.

But I don't see why you think you would get the same results from both procs.PROC SYSLIN is designed to work on simultaneous linear equations. It will run OLS methodologies for each of your three methods. Bear in mind that OLS and 2SLS are essentially single-equation methods, and will not take into account any correlations of errors across equations. Depending on your problem and your data, they may still be prefrable to 3SLS.

However, PROC MODEL is designed to work on non-linear equations and sets of non-linear equations. It minimizes an objective function using a specified goal-seeking algorithm. The two do not use the same methodology to solve the problem. So I don't see that they would come up with the same results. Particularly if the data don't fit the models extremely well.

HTH,
David

谢谢！

我有一本格林的英文版，清华大学出版社的，记不请是第几版了，反正买了有

三四年了，看了一遍被别人借走了，当时也没注意这个问题。

SAS原来也会，还拿到过SAS公司资格证书，但自从5、6年前用上EVIEWS，就再没用过SAS，

现在已基本忘光了

这应该是一个比较新的内容，刚才下载了楼上贴出的用面板数据估计联立方程的文章，是去年写的，竟然有27M，我得好好看，如果哪位高人有现成的程序可以作，希望能共享

我是菜鸟，需要估计如下方程组：
crer   lnsale debtmv regdu1 regdu2 ind52 ind53 ind54 ind55 ind56 ind57 ind58
debtmv crer lnmv roa lnbm regdu1 regdu2 ind52 ind53 ind54 ind55 ind56 ind57 ind58
lnbm crer roa debtmv lnmv regdu1 regdu2 ind52 ind53 ind54 ind55 ind56 ind57 ind58
roa crer lnmv debtmv lnbm regdu1 regdu2 ind52 ind53 ind54 ind55 ind56 ind57 ind58
哪位达人可以帮忙写出用stata做面板数据联立方程的代码。非常感谢：）

用stata 里的reg3，是三阶段最小二乘估计法

用stata软件的reg3命令，可以直接操作