用switchr写了命令来完成转换回归模型,
# delimit ;
eq main : lnwage eduyear1 age age2 nonruralspeci marriage health middle west
eq regime : primary eduyear1 age nonruralspeci marriage health middle west
switchr main regime , noisyn(2064)
结果是如下,问题是
1.它的结果不像是最大似然法估计呢?
2.如果它是最大似然估计的话,那些常见的用ml估计的sigma值如何估算呢?
3 如果它果真没用最大似然估计,如何才能让它用最大似然估计呢?
Here is the regression for the switching eq'n
Source | SS df MS Number of obs = 1629
-------------+------------------------------ F( 7, 1621) =26313.76
Model | 5216.31194 7 745.187419 Prob > F = 0.0000
Residual | 45.9055909 1621 .028319303 R-squared = 0.9913
-------------+------------------------------ Adj R-squared = 0.9912
Total | 5262.21753 1628 3.23232035 Root MSE = .16828
------------------------------------------------------------------------------
primary | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
eduyear1 | -.9784693 .0028502 -343.29 0.000 -.9840599 -.9728788
age | .0374545 .0005546 67.53 0.000 .0363667 .0385423
nonruralsp~i | 1.045758 .0176384 59.29 0.000 1.011162 1.080355
marriage | .8806161 .0138539 63.56 0.000 .8534426 .9077896
health | -1.386456 .0104205 -133.05 0.000 -1.406895 -1.366017
middle | .7592778 .0093303 81.38 0.000 .740977 .7775786
west | -.9105324 .0120501 -75.56 0.000 -.9341677 -.886897
_cons | 9.548603 .0433628 220.20 0.000 9.46355 9.633656
------------------------------------------------------------------------------
On iter 1914 the mean absolute change in the probability vector is : 0.00000
Average of the probability vector is: 0.407
First component regression
(sum of wgt is 6.6346e+02)
Linear regression Number of obs = 1633
F( 8, 1624) = 33.60
Prob > F = 0.0000
R-squared = 0.2135
Root MSE = .7663
------------------------------------------------------------------------------
| Robust
lnwage | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
eduyear1 | .1300825 .0179894 7.23 0.000 .0947977 .1653674
age | -.006009 .0264371 -0.23 0.820 -.0578633 .0458453
age2 | .0000323 .0003137 0.10 0.918 -.0005831 .0006476
nonruralsp~i | -.1530307 .0909135 -1.68 0.093 -.3313507 .0252894
marriage | .116729 .1091475 1.07 0.285 -.0973557 .3308137
health | .0260576 .0594114 0.44 0.661 -.0904733 .1425886
middle | -.6332437 .0566271 -11.18 0.000 -.7443136 -.5221738
west | -.7282977 .0820213 -8.88 0.000 -.8891764 -.567419
_cons | 1.623735 .6129788 2.65 0.008 .421423 2.826048
------------------------------------------------------------------------------
Second component regression
(sum of wgt is 9.6954e+02)
Linear regression Number of obs = 1628
F( 8, 1619) = 52.79
Prob > F = 0.0000
R-squared = 0.2482
Root MSE = .56289
------------------------------------------------------------------------------
| Robust
lnwage | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
eduyear1 | -.0662246 .018213 -3.64 0.000 -.101948 -.0305011
age | .026922 .0163721 1.64 0.100 -.0051908 .0590348
age2 | -.0002956 .0002058 -1.44 0.151 -.0006993 .0001081
nonruralsp~i | -.3702438 .0646978 -5.72 0.000 -.4971441 -.2433435
marriage | -.4167082 .0584343 -7.13 0.000 -.531323 -.3020934
health | .0708556 .0451728 1.57 0.117 -.0177477 .159459
middle | -.4965058 .0374989 -13.24 0.000 -.5700574 -.4229542
west | -.4896273 .0431322 -11.35 0.000 -.5742281 -.4050265
_cons | 3.049916 .3908002 7.80 0.000 2.283389 3.816443
------------------------------------------------------------------------------
(1633 real changes made)
This Switching Regression took 39 seconds.


雷达卡





京公网安备 11010802022788号







