R: st: Tobit regression Model

["Carlo Lazzaro" <carlo.lazzaro@tiscalinet.it>]

Dear Jay,
thanks for the plug.
Yes, I have heard something alike to what you report about handling with
care Tobit model.
Erring on the high side of N is always a good practice. Unfortunately, this
is often a matter of (tight) research funds.
Best wishes,
Carlo


-----Messaggio originale-----
Da: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] Per conto di JVerkuilen
(Gmail)
Inviato: domenica 2 dicembre 2012 17:45
A: statalist@hsphsun2.harvard.edu
Oggetto: Re: st: Tobit regression Model

On Sun, Dec 2, 2012 at 10:57 AM, Carlo Lazzaro <carlo.lazzaro@tiscalinet.it>
wrote:
> As Adel has probably heard during some statistics class, according to 
> a long lasting rule-of-thumb, researcher should have around 20 
> observations for each predictor included in her/his multiple linear 
> regression model. I would assume that this guidance may sound interesting
for Tobit models, too.
> Last time I came across this recommendation was in: Katz MH. 
> Multivariable Analysis. A Practical Guide for Clinicians, 2nd edition.
Cambridge:
> Cambridge University Press, 2006: 81.

For models like tobit, it's probably beter to err on the high side of N.

There are arguments that tobit isn't even a very good model to fit as it is
strongly dependent on the assumption of normality in a way that linear
regression is not. I recall some links to this point posted by Kit Baum that
suggested that a generalized linear model is usually better, but I've not
been able to find them by searching through the archive. I may be
misremembering.
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/