用stata标定离散选择模型

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

stata标定的标定结果为：
为什么没有按选择方案给出参数，数据表中的参数是什么意思啊

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：离散选择模型 Stata 选择模型 tata 是什么意思 模型 选择 Stata 离散 标定

The Coefficients in the table are logit coefficients, i.e. log(odds). Use the command "listcoef", you will get the odds.

我看你还是先找一本stata的书仔细看看把，比如陈峰的现代医学统计与stata应用，论坛上有。

一些基本的东西你应该了解的。

以下是引用woodhaven在2006-12-5 8:11:00的发言：
The Coefficients in the table are logit coefficients, i.e. log(odds). Use the command "listcoef", you will get the odds.

I know how to get the odds after mlogit command. Actually, my question is I cant understand the output (what does the column "coef" meab)after nlogit command and dont know how to get the odds in an NL model (since listcoef doesnt work with models estimated with nlogit ).

Thank you so much for your reply and hope to get your help for the questions mentioned above. Thanks.

Check this link http://www.ats.ucla.edu/stat/stata/examples/greene/greene19.htm

Then go to Greene's book(pay attention to the Version difference), and find the chapter Models with Discrete Dependent Variable. The result table in the books should be the same as the Stata output from the link above.

Since the nested logit model is within the multinomial logit framework, the calculation of odds should be the same. Expoentiating the logit coefficients gives you the odds.

以下是引用woodhaven在2006-12-7 11:48:00的发言：

Check this link http://www.ats.ucla.edu/stat/stata/examples/greene/greene19.htm

Then go to Greene's book(pay attention to the Version difference), and find the chapter Models with Discrete Dependent Variable. The result table in the books should be the same as the Stata output from the link above.

Since the nested logit model is within the multinomial logit framework, the calculation of odds should be the same. Expoentiating the logit coefficients gives you the odds.

楼主帮忙帮到底，我们就拿拿Greene的例子来说吧

nlogit mode (travel = aasc tasc basc gc ttme) (type=hincair), group(grp)
tree structure specified for the nested logit model

top-->bottom

type travel
------------------------
fly air
ground train
bus
car

initial: log likelihood = -207.01917
rescale: log likelihood = -207.01917
rescale eq: log likelihood = -197.65917
Iteration 0: log likelihood = -197.65917
Iteration 1: log likelihood = -197.43396 (backed up)
Iteration 2: log likelihood = -197.18788 (backed up)
Iteration 3: log likelihood = -197.17125 (backed up)
Iteration 4: log likelihood = -197.08544 (backed up)
Iteration 5: log likelihood = -196.84835
Iteration 6: log likelihood = -196.30583
Iteration 7: log likelihood = -194.93294
Iteration 8: log likelihood = -194.86168
Iteration 9: log likelihood = -194.0501
Iteration 10: log likelihood = -193.91879
Iteration 11: log likelihood = -193.75031
Iteration 12: log likelihood = -193.68313
Iteration 13: log likelihood = -193.66092
Iteration 14: log likelihood = -193.65718
Iteration 15: log likelihood = -193.6562
Iteration 16: log likelihood = -193.65615
Iteration 17: log likelihood = -193.65615

Nested logit
Levels = 2 Number of obs = 840
Dependent variable = mode LR chi2(8) = 194.9313
Log likelihood = -193.65615 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
travel |
aasc| 6.042255 1.198907 5.04 0.000 3.692441 8.39207
tasc | 5.064679 .6620317 7.65 0.000 3.767121 6.362237
basc| 4.096302 .6151582 6.66 0.000 2.890614 5.30199
gc | -.0315888 .0081566 -3.87 0.000 -.0475754 -.0156022
ttme | -.1126183 .0141293 -7.97 0.000 -.1403111 -.0849254
-------------+----------------------------------------------------------------
type |
hincair | .0153337 .0093814 1.63 0.102 -.0030534 .0337209
-------------+----------------------------------------------------------------
(IV params)|
|
type |
/fly | .5859993 .1406199 4.17 0.000 .3103894 .8616092
/ground | .3889488 .1236623 3.15 0.002 .1465753 .6313224
------------------------------------------------------------------------------
LR test of homoskedasticity (iv = 1): chi2(2)= 10.94 Prob > chi2 = 0.0042

结果中给出的参数是对于哪一种选择结果的log odds呢，是air,bus,train,还是car，应该如何理解结果列表中的coef列含义，这些参数怎么用？

Since aasc, tasc, and basc are dummy variables of air, bus, train and bus, and the omitted variable is car, we can interpret these logit coefficient in this way. First, exponentiate these logit coefficient and get the odds. Then, you can say, on the first level of the model, after controlling for the effects of the cost and time (gc and ttme) the odds of selecting air versuse travelling by car are ..., the odds of selecting train versus selecting care are ...., and the odds of selecting bus versus car are........

Since the effect of time is negative, you can interpret it like this way: when you increase the time of travel by one minute for any give mode of transportation, the odds of using that mode of travel will decrease by ...., after controlling for the effects of other variables in the model.

You'd better double check with some books, such as Long & Freese (2003), and Powers & Xie (2000). Stata reference does not give too much information, but it will also help you to understand.

I am not an expert in analyzing nlogit modols. If there is any mistake, it is mine.