dtk_yzpbc 发表于 2010-1-17 02:26 
最近在做一个关于农民创业成功率的题目，二元响应变量就是农民创业成功率。由于样本选择的问题，80%的农民创业是成功的，也就是说Y是1，请问这对于模型估计有影响吗？eviews估计发现，常数项为0，这可能吗？恳请高手指点一二。谢谢了
In theory the estimation will be still unbiased for all coefs EXCEPT intercept under the logistic regression assumptions. That is why in many studies in which a sample is stratified on a response variable(y {0,1}). This is the advantage of the logistic regression. But the probabilty estimation will be not true representation of the population.
"常数项为0" is unlikely. It is possible only in certain sampling conditions in this case.
扫码加好友，拉您进群
全部版块
我的主页
收藏
