[PLUG-TALK] metric for mixing/stratification?

[Denis Heidtmann]

On Wed, Jul 18, 2012 at 2:10 AM, Russell Senior
<russell at personaltelco.net>wrote:

>
> I have a set of temperature sensors arranged in a vertical array
> (1x6).  I am looking for a metric I can compute that will represent
> the stratification of temperatures across the array in a single number
> (or just a few).  Something simple like standard deviation would work,
> but it wouldn't take the sensor ordering into account.  Any
> suggestions?
>
> Maybe a linear regression would work, along with an error term.  That
> would tell me the slope in real units (e.g. degrees/meter) which will
> be higher during stratification and lower during mixing.  And the
> error term will tell me how much deviation there is from the straight
> line.
>
>
> --
> Russell Senior, President
> russell at personaltelco.ne <russell at personaltelco.net>t


Can you assume monotonicity?  If not, then the linear regression might not
be the best choice.  Otherwise, the linear regression will provide what you
expect.  If you did a 2nd order fit, you would get additional information
at the expense of more numbers.

If you really want one number, you could report the slope when it is over
some value (stratified), and the SSE when it is below that value (mixed).

-Denis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.pdxlinux.org/pipermail/plug-talk/attachments/20120723/d5ca693e/attachment.html>