没有合适的资源?快使用搜索试试~ 我知道了~
slides_19_twopart.pdf
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 61 浏览量
2022-05-10
09:14:41
上传
评论
收藏 100KB PDF 举报
温馨提示
试读
56页
slides_19_twopart.pdf
资源推荐
资源详情
资源评论
TWO-PART AND HURDLE MODELS
Econometric Analysis of Cross Section and Panel Data,2e
MIT Press
Jeffrey M. Wooldridge
1. Introduction
2. A General Formulation
3. Truncated Normal Hurdle Model
4. Lognormal Hurdle Model
5. Exponential Type II Tobit Model
1
1. INTRODUCTION
∙ We consider the case with a corner at zero and a continuous
distribution for strictly positive values.
∙ Why should we move beyond Tobit? It can be too restrictive because
a single mechanism governs the “participation decision” (y 0 versus
y 0) and the “amount decision” (how much y is if it is positive).
∙ Recall that, in a Tobit model, for a continuous variable x
j
, the partial
effects on Py 0|x and Ey|x,y 0 have the same signs (different
multiples of
j
. So, it is impossible for x
j
to have a positive effect on
Py 0|x and a negative effect on Ey|x,y 0. A similar comment
holds for discrete covariates.
2
∙ Furthermore, for continuous variables x
j
and x
h
,
∂Py 0|x/∂x
j
∂Py 0|x/∂x
h
j
h
∂Ey|x,y 0/∂x
j
∂Ey|x,y 0/∂x
h
∙ So, if x
j
has twice the effect as x
h
on the participation decision, x
j
must have twice the effect on the amount decision, too.
∙ Two-part models allow different mechanisms for the participation and
amount decisions. Often, the economic argument centers around fixed
costs from participating in an activity. (For example, labor supply.)
3
2. A GENERAL FORMULATION
∙ Useful to have a general way to think about two-part models without
specif distributions. Let s be a binary variable that determines whether y
is zero or strictly positive. Let w
∗
be a nonnegative, continuous random
variable. Assume y is generated as
y s w
∗
.
∙ Other than s being binary and w
∗
being continuous, there is another
important difference between s and w
∗
: we effectively observe s
because s is observationally equivalent to the indicator 1y 0
(Pw
∗
0). But w
∗
is only observed when s 1, in which case
w
∗
y.
4
∙ Generally, we might want to allow s and w
∗
to be dependent, but that
is not as easy as it seems. A useful assumption is that s and w
∗
are
independent conditional on explanatory variables x, which we can write
as
Dw
∗
|s,x Dw
∗
|x.
∙ This assumption typically underlies two-part or hurdle models.
∙ One implication is that the expected value of y conditional on x and s
is easy to obtain:
Ey|x,s s Ew
∗
|x,s s Ew
∗
|x.
5
剩余55页未读,继续阅读
资源评论
samLi0620
- 粉丝: 1002
- 资源: 1万+
下载权益
C知道特权
VIP文章
课程特权
开通VIP
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功