求帮忙校对下代码是否有问题

1. local n1 = 100 // number of individuals
   
2. local n2 = 10 // number of periods
   
3. local n = `n1'*`n2'
   
4. set obs `n'
   
5. quietly gen N = 0
   
6. *设定N个 个体
   
7. forvalues i = 1(1)`n1'{
   
8. local k = `n2'*(`i'-1)+1
   
9. local s = `n2'*`i'
   
10. quietly replace N = `i' in `k'/`s'
    
11. }
    
12. *设定panel data期数
    
13. egen T = seq(), from(1)to(`n2')
    
14. 
    
15. xtset N T
    
16. 
    
17. 
    
18. 
    
19. * part II: to obtain unobserved individual-specific effect a
    
20. 
    
21. gen a = 0
    
22. 
    
23. forvalues j= 1(1)`n1'{
    
24. local k = `n2'*(`j'-1)+1
    
25. local s = `n2'*`j'
    
26. local b = invnormal(uniform())
    
27. quietly replace a = `b' in `k'/`s'
    
28. }
    
29. 
    
30. * part III: to obtain Xit.
    
31. 
    
32. * correlation coefficient between X and a: T
    
33. * 解释变量X与个体效应a相关
    
34. local Tao = 0.5
    
35. gen w = invnormal(uniform())
    
36. local t = sqrt(1 - `Tao'^2)
    
37. gen X = `Tao'*a + `t'*w
    
38. 
    
39. * part IV: draw U
    
40. 
    
41. *产生随机扰动项u
    
42. gen u = invnormal(0.5*uniform()+0.5)
    
43. 
    
44. * to generate exogenous variabe z
    
45. 
    
46. gen z1 = rnormal(10,100)
    
47. gen z2 = rnormal(5,60)
    
48. label var z1 "exogenous regressors"
    
49. label var z2 "exogenous regressors"
    
50. 
    
51. * to generate x with measurement e
    
52. gen e = invnormal(uniform())
    
53. gen x = X + e
    
54. 
    
55. 
    
56. * part V: to generate y, X有测量偏误e
    
57. 
    
58. gen y = a + X + z1 + z2 + u
    
59. 
    
60. 
    
61. 
    
62. 
    
63. 
    
64. 
    
65. 
    
66. * ancillary matrix
    
67. 
    
68. matrix one1 = J(`n2',1,1) // to generate a vector whose elements are 1.
    
69. mat one1_t = one1'
    
70. mat A`n2' = I(`n2') - (1/`n2')*one1*one1_t
    
71. mat MR_1 = vec(A`n2')
    
72. mat MR = I(`n2'^2) - (1/(`n2'-1))*MR_1*MR_1'
    
73. 
    
74. 
    
75. mat define M = J(`n2'^2 + 2,`n2'^2 + 2,0)
    
76. mat M[1,1] = MR
    
77. mat M[`n2'^2 + 1,`n2'^2 + 1] = I(2)
    
78. 
    
79. * krocker product of a part of B
    
80. 
    
81. forvalues i = 1(1)`n1'{
    
82. mkmat x if N==`i',matrix(x`i')
    
83. mat x`i'I = x`i'#I(`n2')
    
84. }
    
85. 
    
86. 
    
87. *Zi
    
88. 
    
89. forvalues j = 1(1)`n1'{
    
90. mkmat z1 z2 if N == `j', matrix(Z_`j')
    
91. mat Z_`j'trans =Z_`j''
    
92. }
    
93. 
    
94. * construction of Bi
    
95. 
    
96. forvalues i = 1(1)`n1' {
    
97. 
    
98. mat define B`i' = J(`n2'^2 + 2,`n2',0)
    
99. mat B`i'[1,1] = x`i'I
    
100. mat B`i'[`n2'^2 + 1,1] = Z_`i'trans
     
101. }
     
102. 
     
103. 
     
104. * within transformation
     
105. egen y_meanw = mean(y), by(N)
     
106. egen x_meanw = mean(x), by(N)
     
107. 
     
108. egen z1_meanw = mean(z1), by(N)
     
109. egen z2_meanw = mean(z2), by(N)
     
110. gen dy = y - y_meanw
     
111. gen dx = x - x_meanw
     
112. gen dz1 = z1 - z1_meanw
     
113. gen dz2 = z2 - z2_meanw
     
114. 
     
115. 
     
116. *krocker product of a part of C
     
117. 
     
118. forvalues j = 1(1)`n1'{
     
119. mkmat dx if N == `j' ,matrix(dx`j')
     
120. 
     
121. mkmat dz1 if N == `j', matrix(dz1`j')
     
122. mkmat dz2 if N == `j', matrix(dz2`j')
     
123. 
     
124. mat_capp xz`j'_ : dx`j' dz1`j'
     
125. mat_capp xz`j' : xz`j'_ dz2`j'
     
126. }
     
127. 
     
128. forvalues i = 1(1)`n1'{
     
129. mat C`i'=B`i'*xz`i'
     
130. }
     
131. 
     
132. * construction of dn
     
133. 
     
134. forvalues j = 1(1)`n1'{
     
135. 
     
136. mkmat dy if N == `j', matrix(dy`j')
     
137. mat d`j' = B`j'*dy`j'
     
138. }
     
139. ***************************
     
140. 
     
141. * mean of C and d
     
142. 
     
143. 
     
144. local m = 1
     
145. mat dddd = J(`n2'^2 + 2, 1, 0)
     
146. while `m' <= `n1'{
     
147. mat dddd = d`m' + dddd
     
148. local m = `m' + 1
     
149. }
     
150. 
     
151. mat d_mean = (1/`n1')*dddd
     
152. quietly mat list d_mean
     
153. 
     
154. 
     
155. mat b1 = J(`n2'^2 + 2 , 3 ,0)
     
156. 
     
157. local n = 1
     
158. while `n' <= `n1'{
     
159. 
     
160. mat b1 = C`n' + b1
     
161. 
     
162. local n = `n' + 1
     
163. 
     
164. }
     
165. 
     
166. mat C_mean = (1/`n1')*b1
     
167. 
     
168. 
     
169. 
     
170. * computation of a first step consistent GMM estimator
     
171. 
     
172. * matrix C_mean_trans = C_mean'
     
173. 
     
174. 
     
175. 
     
176. matrix lamda1 = inv(C_mean'*M*C_mean)*C_mean'*M*d_mean
     
177. 
     
178. mat list lamda1
     
179. 
     
180. * its weighting matrix
     
181. 
     
182. matrix dc = J(`n2'^2 + 2,1,0)
     
183. 
     
184. forvalues i = 1(1)`n1'{
     
185. matrix dc`i' = J(`n2'^2 + 2,1,0)
     
186. matrix dc`i' = d`i' - C`i'*lamda1
     
187. }
     
188. 
     
189. mat W1 = J(`n2'^2 + 2, `n2'^2 + 2 , 0)
     
190. 
     
191. local j = 1
     
192. 
     
193. while `j' <= `n1'{
     
194. 
     
195. mat W1 = dc`j'*dc`j'' + W1
     
196. 
     
197. local j = `j' + 1
     
198. }
     
199. 
     
200. mat W_hat = (1/`n1')*W1
     
201. 
     
202. 
     
203. 
     
204. 
     
205. mat lamda2 = inv(C_mean'*M*inv(W_hat)*M*C_mean)*C_mean'*M*inv(W_hat)*M*d_mean
     
206. 
     
207. mat list lamda2

复制代码

附件在另一个帖子中：https://bbs.pinggu.org/forum.php?mod=viewthread&tid=3670626&page=1&extra=#pid29624369

不知为何，附件不可以重复上传！

如上面附件，毕业设计老师让根据这篇文献的内容做一个蒙特卡罗模拟。我的主要内容 是根据文献第三部门的模型做的，即 面板数据的一个解释变量存在测量偏误的情况下，如何通过GMM方法得到有测量偏误解释变量的一致估计量（consistent estimator）。但是，我按照文献中的两步法得到的估计量只有外生变量的系数是一致的，有测量偏误的系数差距比较大。后来我用STATA验证了文献中推导用到的假设，都基本没问题。所以想请STATA高手帮我看看是不是代码存在问题？

谢谢！

* part V: to generate y, X有测量偏误e

gen y = a + X + z1 + z2 + u

这里是大X还是小x？

ywh19860616 发表于 2015-4-19 09:46
* part V: to generate y, X有测量偏误e

gen y = a + X + z1 + z2 + u

谢谢回复！

是大X，因为大X是可以观测到的，小x是不可观测的真实值，e是他们之间的测量偏误。附件的文献就是处理怎么得到有测量偏误系数的一致估计。结果应该是在1附近，但是我编写的程序结果都在0.4左右。