Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: violating normality assumption for tobit
From
"Grant, Robert" <[email protected]>
To
"[email protected]" <[email protected]>
Subject
Re: st: violating normality assumption for tobit
Date
Mon, 27 Feb 2012 11:19:49 +0000
Dear Raymond
I'm not convinced that tobit (or ologit) is the analysis you need because ranks are not the same as grouped / interval-censored values. This is because they are not independent of one another, and some simple non-parametric tests might be better to explore associations with your predictors.
On the other hand if you really do have, as you seem to describe, grouped uniform data then have a look at the papers by Daniel Heitjan on inference from coarsened data, which provide a very general framework, though you will have to do some de novo programming to get any stats package to find such MLEs for you. It's a journey worth making but not a quick one!
Robert Grant
Senior Research Fellow
St George's University of London & Kingston University
020 8725 2281
[email protected]
http://tinyurl.com/helpwithstats
> Date: Fri, 24 Feb 2012 12:43:26 -0500
> Subject: Re: st: violating normality assumption for tobit
> From: [email protected]
> To: [email protected]
>
> Raymond Lim <[email protected]>:
> Meaning, where an ordered probit or ordered logit is clearly preferred.
> help ologit
>
> On Fri, Feb 24, 2012 at 3:07 AM, Nick Cox <[email protected]> wrote:
> > This seems to confuse marginal and conditional distribution. That
> > aside, the way forward here is surely to simulate using what you know
> > to see how -tobit- performs in these non-standard conditions.
> >
> > Nick
> >
> > On Fri, Feb 24, 2012 at 2:37 AM, Raymond Lim <[email protected]> wrote:
> >
> >> I am running a tobit and normality is a crucial assumption. However,
> >> my dependent variable follows a uniform distribution. I'm looking at
> >> award placement: 1st place, 2nd place, ... 10th place. I collect data
> >> from many competitions and always have 10 places. How can I address
> >> the fact that my dependent variable will never be normal even if N
> >> goes to infinity?
This email has been scanned for all viruses by the MessageLabs Email
Security System.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/