Combining uncertainties (error propagation)

In summary: Finally, calculate the uncertainty in x using the above formula:$$\Delta x_a = \sqrt{\left(\frac{\partial x_a}{\partial a}\right) ^2 \Delta a^2 + \left(\frac{\partial x_a}{\partial c}\right) ^2 \Delta c^2}$$In summary, the uncertainty in x is (0.03158 ± 0.00001) / (1.65263 ± 0.00001).
  • #1
Cheesycheese213
55
8
Hi!
I was wondering how would I calculate the uncertainty of a value that is calculated using both multiplication and division?

For example, with something like:

Q = mc(T2 - T1)

I'm not sure what to do with the uncertainties for T2 - T1

uncertainty = Δm/m + Δc/c + (ΔT1 + ΔT2)/(T1 + T2)

Or, would I just leave one of the uncertainties (since the uncertainty for temperature is the same?) so it would be like:

uncertainty = Δm/m + Δc/c + ΔT1 or 2/(T2 - T1)

Thanks!
 
Physics news on Phys.org
  • #2
For what level of course is this? Have you taken any Calculus yet (differentiation in particular)?
 
  • #3
gneill said:
For what level of course is this? Have you taken any Calculus yet (differentiation in particular)?
I have taken calculus, but I'm not sure if we're supposed to use it because we haven't been taught it before for uncertainty? (I'm taking grade 11 physics)
 
  • #4
Cheesycheese213 said:
I have taken calculus, but I'm not sure if we're supposed to use it because we haven't been taught it before for uncertainty? (I'm taking grade 11 physics)
Ah. Okay.

So my recommendation to you would be to divide the process into several steps.

No doubt you've been given formulas for finding the error for the addition/subtraction of two values, the multiplication of two values, and the division of two values.

So perhaps begin by finding the error in the addition of the temperatures. You end up with a value for the temperature difference and its error.

Then take that temperature difference value and associated error and multiply it by m with its associated error, using the appropriate error propagation for multiplication. So now you have a new quantity with associated error.

Rinse and repeat until you've gone through all the terms of the original expression.

The Calculus-based method (which is generally preferred) takes the partial differentials of the original function with respect to each of its variables, then forms a square root of the sum of squares of each partial differential multiplied by the square of the errors. So for example, suppose you had some function of several variables ##f(x_1,x_2,x_3...x_n)## and each of the variables had associated errors ##\Delta x_1, \Delta x_2, \Delta x_3...\Delta x_n##. Then the propagated error in evaluating ##f## would be given by:

##\Delta f = \sqrt{\left(\frac{\partial f}{\partial x_1}\right) ^2 \Delta x_1^2 + \left(\frac{\partial f}{\partial x_2}\right) ^2 \Delta x_2^2 + ... \left(\frac{\partial f}{\partial x_n}\right) ^2 \Delta x_n^2 }##

This works with any expression for ##f##, even it contains square roots, trig functions, or anything else you can find the differentials for.
 
  • Like
Likes Cheesycheese213
  • #5
gneill said:
Ah. Okay.

So my recommendation to you would be to divide the process into several steps.

No doubt you've been given formulas for finding the error for the addition/subtraction of two values, the multiplication of two values, and the division of two values.

So perhaps begin by finding the error in the addition of the temperatures. You end up with a value for the temperature difference and its error.

Then take that temperature difference value and associated error and multiply it by m with its associated error, using the appropriate error propagation for multiplication. So now you have a new quantity with associated error.

Rinse and repeat until you've gone through all the terms of the original expression.

The Calculus-based method (which is generally preferred) takes the partial differentials of the original function with respect to each of its variables, then forms a square root of the sum of squares of each partial differential multiplied by the square of the errors. So for example, suppose you had some function of several variables ##f(x_1,x_2,x_3...x_n)## and each of the variables had associated errors ##\Delta x_1, \Delta x_2, \Delta x_3...\Delta x_n##. Then the propagated error in evaluating ##f## would be given by:

##\Delta f = \sqrt{\left(\frac{\partial f}{\partial x_1}\right) ^2 \Delta x_1^2 + \left(\frac{\partial f}{\partial x_2}\right) ^2 \Delta x_2^2 + ... \left(\frac{\partial f}{\partial x_n}\right) ^2 \Delta x_n^2 }##

This works with any expression for ##f##, even it contains square roots, trig functions, or anything else you can find the differentials for.
Ohhhh ok thank you so much! I think I'll try both of them just in case :D Thanks again!
 
  • #6
Very happy to be of help.
 
  • #7
Here's another method that's similar to gneill's method (actually based on it), but doesn't use any calculus in the calculation. I'll illustrate it with an example.

Suppose x = a / (b + c), where a = 15.4 ± 0.3, b = 5.2 ± 0.2, and c = 4.3 ± 0.3.

First calculate x using the "unvaried" values of a, b, c:
$$x_0 = 15.4 / (5.2 + 4.3) = 1.62105$$
(leaving a few extra figures for now, to prevent roundoff errors)

Then vary (only) a by its uncertainty, recalculate x, and find the difference from the "unvaried" value of x:
$$x_a = 15.7 / (5.2 + 4.3) = 1.65263 \\
\Delta x_a = x_a - x_0 = 0.03158$$
(This is what the uncertainty in x would be, if a were the only uncertain quantity in the calculation.)

Repeat, varying (only) b by its uncertainty:
$$x_b = 15.4 / (5.4 + 4.3) = 1.58763 \\
\Delta x_b = x_b - x_0 = -0.03342$$
Repeat, varying (only) c by its uncertainty:
$$x_c = 15.4 / (5.2 + 4.6) = 1.57143 \\
\Delta x_c = x_c - x_0 = -0.04962$$
Finally, add the differences "in quadrature": square them, add them, and find the square root:
$$\Delta x = \sqrt {{\Delta x_a}^2 + {\Delta x_b}^2 + {\Delta x_c}^2} \\ = \sqrt{0.03158^2 + (-0.03342)^2 + (-0.04962)^2} \\ = 0.06765$$
The result is 1.62105 ± 0.06765. If this is to be reported as a "final" answer, it should be rounded off so the uncertainty has one (maybe two) significant figures: 1.61 ± 0.07.

I hope I haven't made any errors in arithmetic, or in transcription from my calculator. :nb)

Added note: if a quantity appears more than once in the formula, vary it in all locations at once. Don't treat them as separate "variations."
 
Last edited:
  • Like
Likes gneill

1. What is error propagation?

Error propagation, also known as combining uncertainties, is a method used to calculate the overall uncertainty of a measurement or calculation that involves multiple sources of uncertainty.

2. How is error propagation calculated?

Error propagation is calculated by taking the square root of the sum of the squares of each individual uncertainty. This method assumes that the uncertainties are independent of each other.

3. Can error propagation be used for any type of measurement or calculation?

Yes, error propagation can be used for any type of measurement or calculation as long as the uncertainties are known for each variable involved.

4. What is the purpose of error propagation?

The purpose of error propagation is to provide a more accurate representation of the overall uncertainty in a measurement or calculation. It allows for a better understanding of the reliability of the results.

5. Are there any limitations to error propagation?

Yes, error propagation assumes that the uncertainties are independent of each other and that they follow a normal distribution. It may not be accurate for highly complex calculations or when the uncertainties are not normally distributed.

Similar threads

  • Classical Physics
Replies
6
Views
2K
  • Classical Physics
Replies
2
Views
2K
  • Set Theory, Logic, Probability, Statistics
Replies
1
Views
1K
  • Introductory Physics Homework Help
Replies
15
Views
1K
  • Other Physics Topics
Replies
1
Views
2K
  • STEM Educators and Teaching
Replies
11
Views
2K
  • Set Theory, Logic, Probability, Statistics
Replies
22
Views
1K
  • Calculus and Beyond Homework Help
Replies
3
Views
918
Replies
8
Views
805
  • Other Physics Topics
Replies
8
Views
2K
Back
Top