Boost logo

Boost Users :

Subject: Re: [Boost-users] Math/Statistical Distributions
From: Paul A. Bristow (pbristow_at_[hidden])
Date: 2011-03-23 10:30:15


> -----Original Message-----
> From: boost-users-bounces_at_[hidden] [mailto:boost-users-
> bounces_at_[hidden]] On Behalf Of Andrej van der Zee
> Sent: Wednesday, March 23, 2011 8:24 AM
> To: Boost-users_at_[hidden]
> Subject: [Boost-users] Math/Statistical Distributions
>
> Hi,
>
> Statistics have been a while ago... I was reading the documentation, at
least
> part of it, of the Boost library Math/Statistical Distributions.

You should always read ALL the documentation ;-)

>But I think I need a bit of help...
>
> I have many values Yi that are closely distributed around a mean value
Y-mean.
> The collection of Yi is obtained by an algorithm that sometimes produces
trash
> values that need to be discarded. Such false values are typically not
"close" to
> the Y-mean.
>
> I was hoping somebody could tell what distribution I should use and how I
> should use this. If I recall correctly from my school-days, I need the
Normal
> (Gaussian) Distribution. What would be the algorithm to find those trash
values?

I think what you possibly want *first* is an outlier detection (and
discarder?)

http://en.wikipedia.org/wiki/Outlier is a typical intro.

There are distributions (in Boost.Math) that have 'longer tails' - that is
they include what a 'normal' based outlier would throw out.
But if you are trying to predict some 'standard deviation' like measure of
scatter, this probably isn't what you want.

If you just wanted to *display* the scatter a boxplot is useful - outliers
are displayed too.

(A Google Summer of Code project in 2007 produced a working display system
in C++ that you might find useful.

An example is attached.

https://svn.boost.org/svn/boost/sandbox/SOC/2007/visualization/libs/svg_plot
/doc/html/index.html

will give you the manual with some examples. The code is nearby.)

HTH

Paul

---
Paul A. Bristow,
Prizet Farmhouse, Kendal LA8 8AB  UK
+44 1539 561830  07714330204
pbristow_at_[hidden]



Boost-users list run by williamkempf at hotmail.com, kalb at libertysoft.com, bjorn.karlsson at readsoft.com, gregod at cs.rpi.edu, wekempf at cox.net