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]

Subject: Re: st: qnorm

From   "Seed, Paul" <[email protected]>
To   "[email protected]" <[email protected]>
Subject   Subject: Re: st: qnorm
Date   Tue, 6 Mar 2012 10:45:51 +0000

On Mon, Mar 5, 2012 at 4:26 AM, amir gahremanpour wrote:
> In one of my lectures about distributions, I was taught that in QQ-plot all points should be within z=+/-2 !, If we have such a definition we should be able
> to calculate p-value for QQ plot, right? visual assessment of qqplot is very subjective !

It is worth pointing out that significance tests for Normality have a paradoxical property: 
The larger the sample, the more likely you are to get a significant result (due to decreasing SE);
but for large samples, the sampling distribution converges on Normality, non-normality matters less,

For small samples the test lacks power; for large samples it is not needed.

Like Nick Cox I much prefer to use diagnostic plots - they show what the problem is, 
and lead to likely solutions - ignore problem, take logs, drop outliers, or 
recode into categories.  

This should of course be done without regard to the final results - 
it is clearly unsound to carry out the same test in five different ways 
and report only the significant one.  Under some circumstances, 
these decisions might even be made before unblinding the treatment codes.


Paul T Seed, Senior Lecturer in Medical Statistics, 
Division of Women’s Health, King’s College London
Women’s Health Academic Centre, KHP
+44 (0) 020 7188 3642.

"I see no reason to address the comments of your anonymous expert ... I prefer to publish the paper elsewhere" - Albert Einstein

*   For searches and help try:

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