Wallace, John
>
> In Probability and Statistics (JL Devore, Duxbury Press, MA
> 1995) the author
> describes a "Boxplot Rule" for evaluating outliers. Two
> outlier limits are
> calculated using the Interquartile range: the first at 1.5
> * IQR(measured
> symmetrically from the 25th and 75th percentiles), and the
> second at 3*IQR.
> "Mild" outliers, in this scheme are between the two limits
> defined above (at
> either tail). "Extreme" outliers are beyond the 3*IQR limits.
> Note that this is a different application that just a
> pictorial summary of
> the data (which the various plots we've been talking about
> do), the plot I'm
> describing allows you to make judgements about how likely
> it is that suspect
> points in your dataset are outliers. The whiskers of the
> plot would span a
> larger range than the data itself, given a "normal" sample.
> For that
> reason, I would want the whiskers out to 3*IQR replaced by
> the data points themselves.
>
What you want would need some original programming.
I think it is manageable with various tricks,
but by the same token it would be tricky. If
you've not programmed in Stata before, it
would be a formidable first problem.
Nick
n.j.cox@durham.ac.uk
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/