- Jan 21, 2005
- 17,722
- 6
- 81
What would be a good analogy to distinguish precision from accuracy?
- Thread starter Saint Nick
- Start date
shortylickens
No Lifer
- Jul 15, 2003
- 80,287
- 17,080
- 136
I had this once in high school math and a couple times in Navy Advanced Electronics.
One easy analogy is a bulls-eye target.
A bunch of shots around the center is accurate.
A bunch of shots clustered tightly together but away from the center is precise.
A bunch of shots clustered tightly around the center is precise and accurate.
Shots scattered all over the target are neither.
Give me a minute to find some pics.
EDIT: Funny how long it takes to post something huh?
Actually, now that I see my analogy is on Wikipedia I am tempted to rethink it, since Wikipedia is probably the worst damn place to aquire any information.
One easy analogy is a bulls-eye target.
A bunch of shots around the center is accurate.
A bunch of shots clustered tightly together but away from the center is precise.
A bunch of shots clustered tightly around the center is precise and accurate.
Shots scattered all over the target are neither.
Give me a minute to find some pics.
EDIT: Funny how long it takes to post something huh?
Actually, now that I see my analogy is on Wikipedia I am tempted to rethink it, since Wikipedia is probably the worst damn place to aquire any information.
Accuracy is the closeness of a value to the actual value. Precision is the closeness of a group of values to each other. To determine precision, you need more than one value. A group of values may be accurate and precise if they are both close to the actual value and each other.
For an analogy, imagine a target, bullseye, and a marksman with five shots. You can set up a bunch of scenarios with different locations for the bullets to demonstrate accuracy, precision, and accuracy + precision.
For an analogy, imagine a target, bullseye, and a marksman with five shots. You can set up a bunch of scenarios with different locations for the bullets to demonstrate accuracy, precision, and accuracy + precision.
shortylickens' Bull's Eye analogy is a good one. Note one dilemma: to determine accuacy, you need to know the actual answer to a very precise value. But to assess precision, you can use common statistical tools because this parameter is characterized completely by the data itself with almost no external reference.
In the shots-on-target analogy, visually you look at it and judge where the middle of the group of holes is, and the average distance of each hole from the middle. Mathematically, you'd establish a coordinate system, calculate the mean for all holes, then the Standard Deviation for the collection. That is a measure of precision. But you could do all of that with no bulls eye at all - no external reference point. However, you do start to look immediately for one type of external reference - some commonly-accepted measure of what is acceptable precision for this case. That is how you decide whether the precision is good or not. In the shot-on-target analogy, you might ask for an opinion from some "expert" target shooter on how small the group of holes should be, and he / she certainly will ask you one important question before answering: how far was the shooter from the target?
When you want to assess accuracy, you do need external information - in fact, two pieces of information. The most obvious is: is the data average close to the "truth"? In the shot-on-target analogy, the "truth" is established visually - it is the center of the bullseye. Mathematically, the questions translates into: how much (numerically) does the data mean differ from the known "true" value? Often it is hard to establish the "truth". The second related item is: how precisely known is the "truth"? If the external information source can tell you the correct "true" value, it also should provide a numeric estimate of the precision of this value. For example, this calibration weight standard is 2.0040 g, with a Standard Deviation of 0.00073. Given that information, you can use standard statistical tools to decide whether your measurement intrument's data shows it to be within acceptable agreement with the known "truth".
Accuracy often is used to answer the question: can I rely on the answer you give me to be correct, or is there some consistent bias in your results? Precision, on the other hand, tends to be used to answer a different type of question: if your answer today is different from what I expect as a target value, or is different from the data obtained yesterday, should I be concerned or should I ignore the difference as small random variations that are of no significance?
In the shots-on-target analogy, visually you look at it and judge where the middle of the group of holes is, and the average distance of each hole from the middle. Mathematically, you'd establish a coordinate system, calculate the mean for all holes, then the Standard Deviation for the collection. That is a measure of precision. But you could do all of that with no bulls eye at all - no external reference point. However, you do start to look immediately for one type of external reference - some commonly-accepted measure of what is acceptable precision for this case. That is how you decide whether the precision is good or not. In the shot-on-target analogy, you might ask for an opinion from some "expert" target shooter on how small the group of holes should be, and he / she certainly will ask you one important question before answering: how far was the shooter from the target?
When you want to assess accuracy, you do need external information - in fact, two pieces of information. The most obvious is: is the data average close to the "truth"? In the shot-on-target analogy, the "truth" is established visually - it is the center of the bullseye. Mathematically, the questions translates into: how much (numerically) does the data mean differ from the known "true" value? Often it is hard to establish the "truth". The second related item is: how precisely known is the "truth"? If the external information source can tell you the correct "true" value, it also should provide a numeric estimate of the precision of this value. For example, this calibration weight standard is 2.0040 g, with a Standard Deviation of 0.00073. Given that information, you can use standard statistical tools to decide whether your measurement intrument's data shows it to be within acceptable agreement with the known "truth".
Accuracy often is used to answer the question: can I rely on the answer you give me to be correct, or is there some consistent bias in your results? Precision, on the other hand, tends to be used to answer a different type of question: if your answer today is different from what I expect as a target value, or is different from the data obtained yesterday, should I be concerned or should I ignore the difference as small random variations that are of no significance?
mugs
Lifer
- Apr 29, 2003
- 48,920
- 46
- 91
Saying pi = 5.67439892395396548923496543459 is precise but not accurate. Saying pi=3.14 is accurate but not precise.
Edit: I was looking at it from the perspective of a computer scientist, not an engineer: http://en.wikipedia.org/wiki/Precision
Edit: I was looking at it from the perspective of a computer scientist, not an engineer: http://en.wikipedia.org/wiki/Precision
You took the time to calculate the time difference but still managed to screw up the direction?
Viper GTS
depends on how you look at it. If you mean "gain" as in adding a minute and half per day, he's right, just as you "gain" an hour when people with daylight savings wind their clocks back in the fall.
DrPizza
Administrator Elite Member Goat Whisperer
shortylickens
No Lifer
- Jul 15, 2003
- 80,287
- 17,080
- 136
IHateMyJob2004
Lifer
- Sep 29, 2004
- 18,656
- 67
- 91
A basketball player who bounces the ball off the same side of the goal during free throws is precise but inaccurate. A basketball player who bounces the ball off of random parts of the backboard but always into the basket is accurate but imprecise. Nothing but net every time is accurate and precise.
Not sure if the second sentence is true, but could be a good analogy nevertheless with it omitted.
Not sure if the second sentence is true, but could be a good analogy nevertheless with it omitted.
IHateMyJob2004
Lifer
- Sep 29, 2004
- 18,656
- 67
- 91
Ya, because everyone else is doing so damned good.
I QUIT, YOU ALL LOOSE:
http://en.wikipedia.org/wiki/Accuracy_and_precision
TastesLikeChicken
Lifer
- Sep 12, 2004
- 16,852
- 59
- 86
Accuracy is the ability to direct your pee into the toilet bowl.
Precision is doing it without any splashback.
Precision is doing it without any splashback.
Born2bwire
Diamond Member
- Oct 28, 2005
- 9,840
- 6
- 71
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 24K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 19K
-
-
-
Facebook
Twitter