Re: st: Normally distributed error term & testing normality of residuals

[Nick Cox <njcoxstata@gmail.com>]

-qnorm- pays absolutely no attention to where its input comes from.
It's just a variable and there is no adaptation for whether the
variable is a -tobit- residual, nor so far as I can see could there be
any easy adjustment.

This is almost where we started. Residuals from a -tobit- model are
not expected to be normal.

You could simulate according to any data generation process you think
plausible and see what the residuals look like on a -qnorm- plot. I
have no idea how helpful that would be.

Nick

On Sun, Oct 14, 2012 at 9:36 AM, Ebru Ozturk <ebru_0512@hotmail.com> wrote:
> Thank you. Is -qnorm- applicable for Tobit model?
>
> Ebru
>
> ----------------------------------------
>> Date: Sun, 14 Oct 2012 01:15:43 +0100
>> Subject: Re: st: Normally distributed error term & testing normality of residuals
>> From: njcoxstata@gmail.com
>> To: statalist@hsphsun2.harvard.edu
>>
>> As said, I agree that -qnorm- is a good idea. See
>>
>> http://www.stata.com/statalist/archive/2012-09/msg01013.html
>>
>> http://www.stata.com/statalist/archive/2012-09/msg01040.html
>>
>> for why not to choose Jarque-Bera.
>>
>> Nick
>>
>> On Sun, Oct 14, 2012 at 12:09 AM, Justina Fischer <JAVFischer@gmx.de> wrote:
>> > A formal test of normality would be the Jarque-Bera-test of normality, available as user written programme called -jb6-.
>> > It gives nice test stats that can be reported in a paper.
>> > (is there s.th. newer available?)
>> >
>> > I prefer using the Jarque-Bera combined with -qnorm- (see below, which allows to label the single obs) in order to identify outlier observations.
>>
>> Nick Cox <njcoxstata@gmail.com>
>>
>> >> You don't say what kind of model you are fitting.
>> >>
>> >> -qnorm- allows a more incisive graphical assessment of normality than
>> >> -histogram, normal-. It doesn't offer a formal significance test.
>> >> People who are well informed on the matter don't agree: some would
>> >> always seek a formal test, others are happy to think about the results
>> >> of -qnorm- in context.
>> >>
>> >> Even when normality of error terms is an explicit assumption behind a
>> >> model, it is typically also the least important assumption.
>> >>
>> >> Nick
>> >>
>> >> On Sat, Oct 13, 2012 at 12:06 PM, Ebru Ozturk <ebru_0512@hotmail.com>
>> >> wrote:
>> >>
>> >> > I test for normality of residuals with the command below:
>> >> >
>> >> > predict e, resid
>> >> > histogram e, normal
>> >> >
>> >> > So, does this command also work for understanding of whether error term
>> >> normally distributed or not?
>> >> *
*
*   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/