and to how many decimal places?

[fumei]

Just read some text that mentioned a "precise estimation".

Hmmmmm.

I can understand an estimation being as precise as possible, but a precise estimation?

However, I found one definition to be - "conforming strictly to rule or proper form". So if a protocol can only return an estimation, if one follows, strictly, the forms and rules of the procedure, can it really be called precise?

Gerry
My paintings and sculpture

 

[Golom]

In an engineering sense at least, "Precise" and "Accurate" are usually not quite the same thing. For example

"The distance to the moon is 1,207.88916 miles"

is very precise (9 significant figures) ... it's just not very accurate.

The usual assumption is that something is precise if a variation in the last significant digit presented is very small compared to the whole number In the above we would assume an error of estimate of ±0.00005 or about 4 * 10^-6% of the number being specified.

 

[edfair]

When the eyes read "precise estimation" the mind thought SWAG.

 

[fumei]

I understand precise vs accurate. But can "The distance to the moon is about 300,000 kilometers" be called precise? It could be called accurate, because it IS "about" 300,000 km.

If one is estimating, what is variable? The precision, or the accuracy?

Hmmm, maybe I can see the point. If the instrumentation used to measure something has inaccuracies, then the result can be said to be an estimation. But the precision can be high? In that within the context of those inaccuracies you can say the result is precise?

Gerry
My paintings and sculpture

 

[CajunCenturion]

Precision relates to the quality of the process, and accuracy relates to the quality of the result.

--------------
Good Luck
To get the most from your Tek-Tips experience, please read FAQ181-2886
As a circle of light increases so does the circumference of darkness around it. - Albert Einstein

 

[Gooser]

Hmm....

I would say it like this:

Take a look at the tachometer in your vehicle (assumes you have both a vehicle and a tachometer, if you don't imagine one ). The scale on a tachometer usually reads in whole numbers from 1 to about 7, then at the bottom, it says x1000. There may be tick marks between the 1 and the 2, oftentimes there are three tickmarks, dividing RPM into 250rev chunks. Now, if the needle is between 1 1/2 and 1 3/4 one could could assume that your RPM range from 1500 to 1750, not very precise, however if your RPM are within that range, quite accurate.

Now imagine that there are 7 tickmarks between the 1 and the 2, dividing RPM into 125rev chunks. Now, if the needle is just below the 1 5/8 mark, one could assume that the RPM are just under 1625. If this is true, your tachometer is now more precise and more accurate. However, if your needle is just below the 1 5/8 mark but your actual RPM are right at 1630, the guage may be more precise, but less accurate.

Generally speaking, as precision increases, accuracy decreases but this is somewhat subjective. The key is to choose the proper scale for what you are measuring. Don't measure the distance to Cuba in 16ths of an inch. Miles work better, and rounding to the nearest 10 is probably best. 90 miles to cuba is better than 92 mi 3456' 2 7/16". Likewise, when measuring something small, i.e. monitor size, a smaller scale makes more sense, 15.4" makes much more sense than 1' when the 'actual' size is 15.40125". Then, of course, there is alway the nominal size; you don't call a 2x4 a 1 1/2" x 3 1/2" because that's plain dumb. So, the short answer to your question may be, "There is none."



v/r

Gooser

Why do today
that which may not need to be done tomorrow --me

 

[fumei]

The original question was not between accuracy and precision, but whether precision can be bound to an estimate.

Precision relates to the quality of the process

So again, if you take the best measurements possible, even though the result is still an estimation, it can be called precise. Yes?

Gerry
My paintings and sculpture

 

[Trevoke]

Yes.

If you are shooting a gun, and you never hit the center of the target, but you always hit one small general area, then you are a precise shot.
If you are shooting a gun, and you hit the center once, but anywhere on the target the rest of the time, you're neither precise nor accurate.
If you are shooting a gun, and you are always around the center, you're an accurate shot.
If you are shooting a gun and you always hit the center, you're precise and accurate.

Clear?

"That time in Seattle... was a nightmare. I came out of it dead broke, without a house, without anything except a girlfriend and a knowledge of UNIX."
"Well, that's something," Avi says. "Normally those two are mutually exclusive."
-- Neal Stephenson, "Cryptonomicon"

 

[sbudzynski]

If you have a formula that guesses future growth based on past growth history then that is an estimate. The further back in history you collect your data then the more precise the estimate. The accuracy cannot be known until the event occurs.

----------

Steve Budzynski


"So, pass another round around for the kids. Who have nothing left to lose and for those souls old and sold out by the soles of my shoes"

 

[CajunCenturion]

==> but whether precision can be bound to an estimate.
Yes, but not in the sense that it means the estimate is accurate, but yes in the sense that you used a precise method to obtain the estimate.

If you take the best possible measurements, then yes, you have a precise measurement, but if you're measuring the wrong thing, the estimate won't be very accurate.

If you use a 24 digit calculator to estimate the product of half a dozen numbers, you'll have a very precise estimate, but if one of your numbers is wrong, the estimate won't be very accurate, even though the process is very precise. On the flip side, if you use a slide rule to estimate that same product, you can get a very accurate estimate, (especially if you use the right numbers), even though the process is not very precise.

That being said, 50 years ago, one could argue that the slide rule was rather precise, and at that time, perhaps it was.

--------------
Good Luck
To get the most from your Tek-Tips experience, please read FAQ181-2886
As a circle of light increases so does the circumference of darkness around it. - Albert Einstein

 

[Gooser]

... if you take the best measurements possible...

What do you mean by 'best'? What is good about the measurement? Good accuracy, or good precision? Or both?

I do believe that this is an accuracy v. precision argument in that for an estimate to be 'precise', you would have to, no doubt, sacrifice some accuracy. For example, if I say that my next birthday will fall precisely on 1 January 2007, that is very precise (on a cosmic scale) but horribly inaccurate since that is not my birtday. If, however, I loosen the scale (and with it the precision) to weeks, and I say that my birthday will fall in the first week of January 2007, then I am now accurate because my birthday falls on 6 January.

IMHO any time you talk about precision OR accuracy you MUST speak to precision v. accuracy.



v/r

Gooser

Why do today
that which may not need to be done tomorrow --me

 

[CajunCenturion]

I don't see the two at odds since precision qualifies the process and accuracy qualifies the result.

==> in that for an estimate to be 'precise', you would have to, no doubt, sacrifice some accuracy.
I don't follow that line of reasoning. Let's assume that I need to estimate the distance between two points, and I have three methods available. I can step off the distance, I can use a tape measure, or I can use a laser range finder. The first method is the least precise process and will most likely yield the least accurate result. The tape measure is more precise than counting steps, and will probably result in a more accurate answer. The third process is the most precise method and should result in the most accurate result.

--------------
Good Luck
To get the most from your Tek-Tips experience, please read FAQ181-2886
As a circle of light increases so does the circumference of darkness around it. - Albert Einstein

 

[fumei]

Let me make the context. I am reading a first edition book by the Chief Medical Officier of the 1904 Younghusband Expedition to Tibet. The book was published in 1905. It contains (supposedly) the first ever colour photograph ever published. This was the first (and ONLY) army to cross the Himalayas from south to north. They also did it in winter! They actually had battle engagements at altitudes of > 16,000 feet. Flanking Goorka units fought as high as 19,000 feet. No oxygen of course.

Temperatures as cold as -25 F (which was as low as their instruments went) - no fuel, no heat.

He was also a founding member of the Geological Survey of India, and a very asute observer. He was part of the team to establish that Everest was the highest peak.

As the Expedition moved north, he and others of the scientific part of the army did a lot of surveying. One of the things they did a lot of (naturally) was to get altitude readings. Of the passes they went through. Of lakes and rivers, etc. etc. etc.

He mentions one day the wind was so turbulent that they "had difficulty getting precise estimates" of altitudes.

He also mentions a "boiling water barometer". Anyone know what that is?

Gerry
My paintings and sculpture

 

[CajunCenturion]

==> He mentions one day the wind was so turbulent that they "had difficulty getting precise estimates" of altitudes.
To more wind that you have, the harder it is to get a good measure of air pressure. The harder it is to get an accurate pressure reading, the harder it is to determine altitude.

A boiling water barometer is a device that is used to determine the barometric pressure by measuring the temperature at which water boils. The less the air pressure, the lower the boiling temperature of water. By knowing your altitude, and the standard pressure and boiling temperature at that altitude, you can determine the real air pressure by measuring the actual temperature at which water boils and comparing it to the standard boiling temperature for that altitude.

--------------
Good Luck
To get the most from your Tek-Tips experience, please read FAQ181-2886
As a circle of light increases so does the circumference of darkness around it. - Albert Einstein

 

[Gooser]

Ah, but you are comparing the accuracy of one to the accuracy of another--I'm not sure whether this is valid.

Isn't the accuracy a measure of whether you are within the limits of your unit of measure?

Now you are confusing me!

precision- the degree of refinement with which an operation is performed or a measurement stated

accuracy- degree of conformity of a measure to a standard or a true value

--source m-w.com - I chose the definitions that best suited my argument, admittedly one of the definitions of precision includes the word accuracy which further muddies my glasses.

The way I understand it, given a value, depending on the precision, it varies whether or not you will be 'accurate' that is, within the limits of your measure.

The plot thickens.

v/r

Gooser

Why do today
that which may not need to be done tomorrow --me

 

[CajunCenturion]

precision- the degree of refinement with which an operation is performed or a measurement stated

accuracy- degree of conformity of a measure to a standard or a true value

Isn't that what I've been suggesting all along. Precision qualifies the process (operation), and accuracy is qualifies the value.

--------------
Good Luck
To get the most from your Tek-Tips experience, please read FAQ181-2886
As a circle of light increases so does the circumference of darkness around it. - Albert Einstein

 

[fumei]

Ok...so, even though it is recognized that taking these barometric measurements can only give an estimate, if on:

Day A: it is REALLY windy and therefore hard to get an accurate reading, the estimate is LESS precise than:

Day B: when the wind is calm.

If precision is - "the degree of refinement with which an operation is performed", can it be said Day A is in fact less precise? It is the reading achieved. It is the best the instrument was capable of giving. Less accurate...sure, but less precise?

And I still have problems with ANY estimate being precise. If he had stated "due to high winds the accuracy of our altitude readings was reduced.", would that not be more....accurate?

Gerry
My paintings and sculpture

 

[Gooser]

fumei:Maybe they take three readings and average...

That way if the three readings were further apart they'd be less precise?

Maybe he just misspoke.

Centurion: I guess I was looking at the OR...

...or a measurement stated...

As in, the degree of refinement with which a measurement stated.

I've always been told (I was previously a draftsman, among other things) that precision has a direct and opposite correlation to accuracy.

Take for example a steel column in a structure, the target (calculated) height is 23'-2.1375". Now, say this steel column must be cut from a I-shaped beam by a guy with a torch, we'll call him Bubba. Then a baseplate and a cap must be welded to it at the bottom and top, respectively. Of course there is no way in he11 that he will be able to get it exactly 23'-2.1375". So, how close does it need to be to the calculated height (i.e. how accurate)? How precise should the column be dimensioned on the shop drawing that he will use to build it? If he has to make 100 of these columns, how many scrap columns are okay? 1 in 10, 1 in 25, 1 in 50, or none. The more precise a measure we hold that dimension to, the harder it is going to be for Bubba the happy fabricator to hit that target. If we make the dimension 23'-2.1375" that gives us a range of 23'-2.13745" < x < 23'-2.13755" a range of less than 1 ten-thousanth of an inch. Leaves you with a big pile of scrap columns. If you dimension it to 23'-2-1/8" that gives you a range of 23'-2-1/16" < x < 23'-2-3/16", still hard to hit, but the pile of scrap is shrinking. If you dimension it to 23'-2-1/4" that gives you a range of 23'-2-1/8" < x < 23'-2-3/8", still shrinking that pile of scrap. If you dimension it to 23'-2-1/2" now the range 23'-2-1/4" < x < 23'-2-3/4", these columns are getting easier and easier to make, but uh-oh look what's coming. If you dimension it to 23'-2" now the range 23'-1" < x < 23'-3". Now when Jim-Bob goes to line these things up with the rafters as he is erecting the building, the damn things don't line up. So, what do you do? You have but one choice, find the perfect balance between precision and accuracy.

v/r

Gooser

Close only counts in horeshoes, hand-grenades and tactical nuclear weapons.

 

[CajunCenturion]

Yes, I think it can be said that the measurements are less precise on Day A because the conditions prevent the instruments from operating at peak efficiency. The instruments cannot be as precise in those conditions as they would be on a calm day.

However, that doesn't mean the result of the measurement is less accurate. It probably will be, but not necessarily so. From my previous example, the answer I get from counting steps may actually be exact, even though that process is far less precise than using a laser range finder.

==> And I still have problems with ANY estimate being precise
And you should. "Precision" is not the metric used to determine the "correctness" of an estimate. Precision is the metric used to determine how refined the process is that was used to obtain the estimate.

Accuracy is the metric used to determine the correctness of the estimate.

--------------
Good Luck
To get the most from your Tek-Tips experience, please read FAQ181-2886
As a circle of light increases so does the circumference of darkness around it. - Albert Einstein

 

[Gooser]

Well said, CajunCenturion!

v/r

Gooser

v/r

Gooser

Why do today
that which may not need to be done tomorrow --me