Experimental Standards?

Why not Google 'International Scientific Standards'?

It would be interesting to find out about.

Let us know what you find. or.

I, of course, am too lazy to look it up myself.

Basic experimental design:

1. The basic scientific method is to test different values of one experimental variable, while all other variables are kept constant. This allows you to collect meaningful data about the effects attributable to your experimental variable.

2. In designing the experiment, you have to consider possible sources of error. Your setup should include a plan to mitigate any identified sources of error: this method is written up in your paper.

3. Statistical methods are built into the experimental design, in order to determine your margins of error and to show how stable your setup is and how reproducible your results are expected to be. So, for example, instead of taking a measurement once, you collect a data set of 30 or more measurements of the same test, and so on for each value of the experimental variable.

4. Your data should be collected in writing (digitally nowadays, in some permanent format), and should be kept not only to perform analysis, but to back up your analysis if it is challenged. Where applicable, physical data should also be retained.

These general procedures apply to any branch of science. If you discuss your specific research project, or ask about any of these design procedures, I'm sure you will get good advice about each or any of these issues.

Did I leave anything out?

Perhaps a bit of elaboration on the subject of designed experiments. It's a statistical analysis subject I have some overarching awareness of but certainly not enough thorough knowledge to attempt an explanation. I'd sure like to read more basic commentary on the subject; but don't want to narrow the topic at this point. (I have a book on the subject; but these days I tend to fall asleep before I get more than a page or two into the second chapter of most books which start the math at about that point)

Hmmm........please don't answer this for a couple of days.....

Ed Weldon

It is not always necessary to make 30 measurements. this number was chosen to be sure that the equations for the normal distribution (Gauss) can be used since in fact they are valid only for an infinite number of elements in the sample.

There is an other approach called the Student method which permits to deal with smaller sets of values and which has as limit the Gauss distribution if the sample is big enough.

Depending on the limits of uncertainty the size can be reduced to less than30.

Good point. Much depends on the specific objective. I personally have a great interest in "small n" research. Sometimes it is the only practical approach: sometimes the necessary knowledge can be obtained with a small sample.

If the objective is to have your findings published and recognized world-wide, the stronger statistical basis is the way to go. Then again, the statistical expectations vary from one branch of science to another. I know nothing about fluid dynamics or the standard statistical approaches to this type of experiment. If the setup is really top notch, maybe there's little variation in the measurement taken, and no need to repeat each measurement so many times. So the student variable would be fine.

For general ideas about experimental design standards, I also found this page, (although the examples are medical the general outline is pretty comprehensive)

http://www.okstate.edu/ag/agedcm4h/academic/aged5980a/5980/newpage2.htm

here's one that uses standard experimental designs
http://www.camo.com/rt/Resources/design_of_experiment.html

I don't know, try here, here and here.

but I'm slowing building a new application for a genre that I'm calling "Computer Aided Specification and Compliance". I need a foundation specification to work with, so if you are going to write one (sounds like a great project), I would happily incorporate it as a base module, and you would have a bit more marketing for your work.

Chris

I just re-read the opening question, and note the the application is Fluid Dynamics. WOE TO YOU!!!

If your flow is laminar, there is some hope. If there is significant turbulence, it will be nearly impossible to make measurements that are sensible in the spirit of my previous comment. Good Luck!

Dear drobertson,

Well said... however, what I think you have really done here is define the difference between an Engineer and a Scientist.

A scientist 'should' do exactly as you say, as having an experiment, with a corresponding theory, with reproducible results, is the purpose scientific method.

An engineer curve fits continously, because they are all about figuring out the best use of natural and real effects, and putting it to work. So they will modify and tweak all over the place.. and if the theory goes out the window, oh well. just have to come up with a new theory. In the meantime, "how cool is this new gizmo I just engineered!"

Chris

"So they will modify and tweak all over the place.. and if the theory goes out the window, oh well..."

OOPS there goes another Rubber Tree Plant, or bridge, or reactor, or ..........................

I agree completely. I happen to like being both at different times. There are abut 7 different projects in my garage that are in different states of "Tweak". Personally I enjoy the thrill of tweaking a design until it works just the way I want. It is a very satisfying process.

As for the answer above I was just trying to make a point relative to the thread title "Experimental Standards". He seems to be questioning from a more scientific point of view and hoping to get results that are credible to outside people. A scientific process is better for him in that context. On the other hand, when the boss wants something last week the scientific process goes out the window. Tweaking and gut instinct take over and you get it done.

-Doug

I am doubtful, if there is any international standard for this subject. But, there whole science for it which is called Design of Experiment. This goes on ellimil\nating uncertainties of various parameters and goes of defining effect of each parameter on functioning of object user study