Why is one significant figure the norm in uncertainties?

367
13

Main Question or Discussion Point

Why is it that only one significant figure for any uncertainty is taught? It seems like a nice general rule, but isn't it an unnecessary constraint and ultimately a poor rule in many cases? For example ## 5 \pm 1 ## could mean an error of either 0.5 or 1.5, which is a very large range. Wouldn't it be best to keep an additional significant figure in this case? If not, could you possibly expand on why? I have also seen more than one significant figure widely reported in research, and it doesn't seem to be too problematic in publications, yet I see this commonly taught at the high school and undergraduate level. Any advice would be great!
 

Answers and Replies

ZapperZ
Staff Emeritus
Science Advisor
Education Advisor
Insights Author
2018 Award
35,222
4,039
Why is it that only one significant figure for any uncertainty is taught? It seems like a nice general rule, but isn't it an unnecessary constraint and ultimately a poor rule in many cases? For example ## 5 \pm 1 ## could mean an error of either 0.5 or 1.5, which is a very large range. Wouldn't it be best to keep an additional significant figure in this case? If not, could you possibly expand on why? I have also seen more than one significant figure widely reported in research, and it doesn't seem to be too problematic in publications, yet I see this commonly taught at the high school and undergraduate level. Any advice would be great!
Er... ## 5 \pm 1 ## means that the value could be in the range of 4 to 6. Not sure where you get that the error is ".. either 0.5 or 1.5.." here.

There is no set standard for uncertainty. It depends on the precision of the measurement or instrument. Maybe what you are being taught is the concept of uncertainty and errors, rather than an actual measurement. So to make it easier to understand the concept, you are being taught simple stuff. And having just one significant digit in the uncertainty looks simple to me.

Zz.
 
Vanadium 50
Staff Emeritus
Science Advisor
Education Advisor
23,453
5,882
You are right that there is a big gap between ±1 and ±2, although there is not such a big gap between ±0.9 (also one digit) and ±1. Some people use more complex rules, for example, here is what the Particle Data Group does:

The basic rule states that if the three highest order digits of the error lie between 100 and 354, we round to two significant digits. If they lie between 355 and 949, we round to one significant digit. Finally, if they lie between 950 and 999, we round up to 1000 and keep two significant digits. In all cases, the central value is given with a precision that matches that of the error.
In your example, they would use 5.0 ± 1.0.

The problem they are trying to avoid is a case like 500 ± 101. With an error of 101 and not 100, we are saying we know the uncertainty to about a percent. However, we only know the central value to 20%. It is highly unlikely we will know the uncertainty 20x better than the central value.
 
367
13
Er... ## 5 \pm 1 ## means that the value could be in the range of 4 to 6. Not sure where you get that the error is ".. either 0.5 or 1.5.." here.

There is no set standard for uncertainty. It depends on the precision of the measurement or instrument. Maybe what you are being taught is the concept of uncertainty and errors, rather than an actual measurement. So to make it easier to understand the concept, you are being taught simple stuff. And having just one significant digit in the uncertainty looks simple to me.

Zz.
I guess I was just trying to say: relative to the error (i.e. 1), if the known error is actually 1.4, then there is quite a big difference between ## 5.0 \pm 1 ## and ## 5.0 \pm 1.4 ##. This appears to be magnified if instead of 5.0, another lower measured value was found (e.g. 1.5 or 0.9), and in those cases, the realtive error caused by using either 1 or 1.4 is quite significant. This could just be easily resolved if we let the error simply be its true value, instead of doing further rounding, from what I see.

Fair enough.
 
367
13
You are right that there is a big gap between ±1 and ±2, although there is not such a big gap between ±0.9 (also one digit) and ±1. Some people use more complex rules, for example, here is what the Particle Data Group does:



In your example, they would use 5.0 ± 1.0.

The problem they are trying to avoid is a case like 500 ± 101. With an error of 101 and not 100, we are saying we know the uncertainty to about a percent. However, we only know the central value to 20%. It is highly unlikely we will know the uncertainty 20x better than the central value.
Thank you. Just to be clear, although unlikely, it is still possible to know the uncertainty better than the central value, right?
 
Vanadium 50
Staff Emeritus
Science Advisor
Education Advisor
23,453
5,882
Possible? Sure. Likely? No. I can't think of any time I have had that happen.
 

Related Threads for: Why is one significant figure the norm in uncertainties?

Replies
2
Views
475
Replies
13
Views
949
Replies
8
Views
659
  • Last Post
2
Replies
32
Views
2K
  • Last Post
Replies
4
Views
3K
  • Last Post
Replies
6
Views
6K
  • Last Post
Replies
5
Views
2K
Top