[Stoves] Fuel test procedures

[AJH]

Forwarding a reply Steve Taylor made to my earlier questions, AJH

On Mon, 09 Oct 2006 22:04:58 GMT, Steve  wrote:


> I'll join in if it's simple but I've never been much good
> at obeying ;-)

You don't have to, but when it comes to the stats analysis,
you can usually spot the folks playing fast and loose !

> 
> Sound pretty auspicious Steve, tell us more. As it is the
> American standards body can we agree on units other than
> seconds?

All ASTM methods are in SI now AFAIK
 
> I flunked statistics so don't understand the significance
> of regressions, any chance of a simple guide?

There is a lot more to it than regression - I am not a
statistician, but an engineer,but statistics are closely
related to probability as well. The basic aim of an ASTM
method is to define REPRODUCIBILITY  - that's the ability of
different rigs with different operators to agree most of the
time with a specified percentage, and the REPEATABILITY,
that is two different tests on the same rig by the same
operator  agree within how much. Typically, for the method
related to one of my instruments for example, the
reproducibility is ~2.5% and the repeatability is  <1 %.
Statistically speaking, I am assured, a ratio of 2:1 in 
those two figure is good. Much bigger than that, and the
stats are somewhat suspect. There is one method I know of
which had a repeatability of 2-3% but a reproducibility of 40%
- tossing dice was almost as accurate a predictor of a test
result as the test....

Process wise, ASTM standards are what are called "consensus
standards"- everyone with a stake in the standard is free to
be involved in the development of the standard and   to
criticise/dismiss the results. In my professional case, as
an instrument maker, I face a committee of users (oil
companies and car manufacturers) and competitor makers. 

A member of the committee which oversees a test
method/standard may affirm, abstain or object to a vote on
it, but in the event that they object MUST explain on
exactly what grounds they make their objection - so you
can't vote negative because it wasn't your idea or
product....  The negative vote can be ruled "persuasive" or
not. A persuasive negative is a return to square one, and
more analysis of why the test method proposed failed to work
properly. Because the membership of the committees are
totally open, and large, the general result are good methods
which are clear to use and give reproducible answers - which
it seems to me is something the group is looking for. 

The usual procedure is to arrange a group of people with the
same equipment, and a small range of test samples (fuels for
us), usually blind coded. The agreed test method is
carefully applied by all the users and the results on all
the sample data returned to be analysed. Usually in ASTM the
lab names, and all the test samples are munged, so no-one
can draw inferences from a given lab's results. 

Since I have the experience with the processes, I am happy
to offer to do the necessary work to try the idea out, but I
have next to no experience with practical stoves, or
prototype fuel standards, and I would need a lot of input
from fellow list members. 

Hopefully at the end of a (short) process, we will be able
to make sound inferences as a result of applying our own
agreed tests.


Steve