Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: zero-inflated Tobit models

From   Ana Paula Fernandes Jubran <>
To   "" <>
Subject   Re: st: zero-inflated Tobit models
Date   Mon, 10 Jun 2013 02:43:50 +0100

Please, I would like to suspend my subscription. Thanks.

Sent from my iPad

On 10 Jun 2013, at 02:18, Austin Nichols <> wrote:

> Katie Farrin <>:
> From what you say, your data is not censored at zero; you observe
> negative profits when they occur.  Instead you have a two-part model,
> where the probability of owning a business is driven by one process
> and net profits conditional on owning a business by another (though
> you would expect that some of those reporting negative profits may
> shortly be transitioning to not owning a business).  There is no
> reason to run a tobit, zero-inflated or otherwise.  You might consider
> an SEM or GMM model that combines the selection into operating a
> business with the profit equation, allowing for the correlation of
> error terms in the two models.  Or wait a couple of weeks and try:
> On Sun, Jun 9, 2013 at 3:42 PM, Katie Farrin <> wrote:
>> Hi, Statalisters,
>> I'm working with data with left-censored values (at zero) and am thus
>> far using a Tobit model.  One of my professors mentioned a
>> zero-inflated Tobit versus a regular Tobit.  My dependent variable,
>> business profits, is censored at zero for survey respondents who do
>> not own any part of a private business.  However, for those who do
>> report business income (only about 1/5 of the sample), profits can be
>> negative.
>> First, I am wondering if there is a Stata program that performs
>> zero-inflated continuous data regressions.  I have not come across a
>> program that does zero-inflated Tobit.  Second, I was wondering if
>> anyone familiar with the models could give a brief explanation of the
>> advantages of a zero-inflated model compared to a standard Tobit; I
>> have read articles showing differences in parameter estimates using
>> the two types of models, but am a little confused about the
>> assumptions regarding the data generation process when I compare the
>> two models for my research.
> *
> *   For searches and help try:
> *
> *
> *
*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index