欢迎投稿(荐稿)计量经济圈,计量相关都行
邮箱:econometrics666@sina.cn
编辑:计量经济圈;转载注明出处。
- Truncated regression
Truncated regression与Censored regression有一些不同,其区别在于前者不仅Y看不见,对应的X也看不见了,而后者虽然归并了Y,但是对应的X能够看见。Tr的样本数实际上是减少了,而Cr样本数实际上并没有减少。而且Cr是归并Y,而Tr不仅可以依据Y来删除样本,而且有时候依据X来删除样本也可能造成OLS的有偏估计。
两者使用的估计方法都是ML,两者都假定那个扰动项error服从Truncated normal分布,从而Tr依据条件分布概率来构造似然函数,而Cr把似然函数分成两部分,即p(y*>0) (假设左归并)和p(y*=0),都构造了一个逆米尔斯比例 λ。
在Stata中分别使用truncreg和tobit命令来完成回归。想要大致了解Truncated regression,可以看看下面附上的图片,画黄颜色的部分比较重要。
- Censored regression
归并回归(censored regression)的特点是对于某些观测数据,被解释变量被压缩到一个点上了。此时被解释变量的概率分布就变成由一个离散点与一个连续分布所组成的混合分布,用OLS来估计,无论使用的是整个样本,还是去掉离散点后的子样本,都不能得到一致的估计。
Tobin(1958年)提出用MLE估计这个模型,该方法因此被称为“Tobit”,也称为“Type I Tobit”或归并回归(censored regression)。伍德里奇的横截面与面板数据的经济计量分析的16章以及陈强老师的书里面的14章有详细的分析。
Tobit estimation was originally developed by Tobin (1958). A consumer durable was purchased if a consumer’s desire was high enough, where desire was measured by the dollar amount spent by the purchaser. If no purchase was made, the measure of desire was censored at zero.
xttobit fits random-effects tobit models(这是用于panel data的tobit命令). There is no command for a parametric conditional fixedeffects model, as there does not exist a sufficient statistic allowing the fixed effects to be conditioned out of the likelihood.
Honor´ e (1992) has developed a semiparametric estimator for fixed-effect tobit models. Unconditional fixed-effects tobit models may be fit with the tobit command with indicator variables for the panels; the indicators can be created with the factor-variable. However, unconditional fixed-effects estimates are biased.
注:文章来源于以下打卡项目。
点击下图,有惊喜