Ken, I assume that you're a mathematician.
I’m a mechanical engineer, with a fair amount of experience as designing, running & reporting on experiments to both outside companies (Boeing, NASA, the FDA) and much more often, to the engineering department within the companies for which I worked. The majority of those experiments & reports ended up defining, or changing, the direction that various projects were headed. They needed to be correct.
My overall comment is “settle down, cowboy.”
You ain’t the only smart guy in the room.
Do you usually find this aggressive sort of approach to be productive on a first introduction?
Or do you get punched in the face periodically?
__
I'm not sure if the 0.2 there is correct.
Neither am I. That would take a boatload of additional experimentation.
But it is
justifiable based on my SMALL SAMPLING of experimental data. I tested 4 balls, of which I could use 3 complete data sets, because the 4th sprung a leak late in the testing cycle. (But it was patently obvious that the 4th ball was following EXACTLY the same trendiness as the other 3. Just bad form to use data when you don’t know for certain when it started to deviate.)
So, as Dr. Pain says, he doesn’t need something that’ll be traceable to NBS Standards. (Showing my age. Is that NIST standards, these days? ISO 9000?)
Engineers like answers. Not questions.
The ±0.2 psi is my sense of the consistency & reliability of the number that pops out of the app, without reference to the error of the input data.
it is intended to give a FEELING to the person reading it of my sense of the error bands likely in the calculation.
It may well be that ±0.3 or ±0.5 turns out to be the case, if you’re trying to get 3 sigma or 6 sigma confidence interval.
But who cares.?
Read what Dr. Pain asked for.
The number that I suggested is adequate to give someone an idea, a feeling for both the number & the precision of the calculation.
__
Ideally this is done using what's called "interval arithmetic" but here there are several simpler options, and most scientists themselves don't use interval arithmetic, clearly overkill for this problem.
WTF??
There is no need for “interval arithmetic” in any of this.
The simple empirical equation that I gave him IS a “simpler option”.
__
Just think about how it's used.
Jeebus. Thanks, Ken. I’ll try...
__
First, I would consider just making a disclaimer about precision.
No. A disclaimer transmits NO information about error or precision.
Providing an error band gives one a SENSE of the precision of the calculation.
That is all that I’m trying to do.
This is the SENSE that I got from running the experiment & observing the results.
__
Second DO NOT TRUNCATE as Tom Kordis above suggests. This is utterly wrong. Use rounding if anything.
You got me. Wrong word.
Don’t truncate, round off.
__
You have to understand why things are what they are you see.
PV = nRT ain’t no big mystery, Ken.
And NOTHING in what Dr. Pain asked for requires this level of detail or complexity.
__
Let's say you want to convert a psi delta of 2 into Fahrenheit. Where is the "2" here from?
Well, Dr. Pain asked for P2 = f(P1, T1, T2), and that’s what I gave him.
He didn’t ask for T2 = g(T1, P1, P2).
But, if he had, then I would have given him this one:
T2 = T1 + ((P2 - P1)/PLR)
with a suggested error band that I’ll work out, when I don’t have to be somewhere, with much more fun, interesting people.
__
Maybe "2" is a mathematical construct, exact, someone is asking a question about physics. In this case, you can use an exact value.
But if "2" comes from a measurement, then maybe "2" is due to rounding of "between 1.5 and 2.5. In this case, the result should be given as a range of temperature between the temperatures corresponding to exact PSI deltas of 1.5 and 2.5.
Thanks, guy.
I understand very well error analysis & precision of measurements.
Last time: READ WHAT Dr. Pain ASKED FOR.
__
This is all utterly obvious if you think about it. In fact, if it's not obvious to you, don't implement. If you can't see why "truncating to 1 significant digit" is wrong, for instance, then you're not understanding the issue.
Please, down boy.
I’ve been doing experimental analysis for every company that I’ve ever worked for for over 40 years.
It IS all “utterly obvious”.
You got me. Round off. Don’t truncate.
By the way, I did NOT say "truncating to 1 significant digit”.
(I should have said) “Round to" (and I did say) "
1 digit after the decimal point.”
IN THE RANGE OF PRESSURES that we’re discussing, that give either 2 or 3 significant digits. With an error band.
By the way, I HAVE run these experiments.
Have you?
As a result, I DO have a feeling for these numbers.
Do you?
__
Your Conclusions:
[QUOTE="Ken Canin, post: 4178924, member: 35283”]
A. You need to include something about significant digits or the use of the program with uncertain input data...
B. Consider optionally outputting the answer as a range if the input data (say pressure) is also a range...
C. Generally if you're kind of trying to learn about apps, it's best to get the basics right first...
[/QUOTE]
__
My conclusions:
Sometimes, smart guys can be insufferable, pompous assholes.
Just my opinion, of course…
YMMV.