I am currently interested in learning more on error propagation. At the moment I am trying to find out how to calculate the uncertainty of a value that is obtained from a linear model. For the linear model, the uncertainty of the slope and of the intercept with the x-axis are given. I have not found any useful explanation or I did not recognize one, because they were too complex for me. The Model would be:

y = s * (x-X0)


s=slope (=0.9)
x=value on the x-axis (=25)
X0=intercept with the x-axis (= 5)

I want to know dy, the uncertainty of y. I know that:

ds = uncertainty of s (= 0.01)
dx = uncertainty of the actual x-value (= 2)
dX0 = uncertainty of the intercept with the x-axis (= 1.5)

This appears to me as a simple problem, so it should not necessary to solve the model with all possible values s, x and X0 could have. Or am I wrong here? I also appreciate recommendations on text-books, if you have any. Kind regards, S.


Yes you are correct.

For other readers, there are a number of clear summaries of how to propagate errors through a linear system.

Here are three of the best ones I found:

  1. http://ipl.physics.harvard.edu/wp-uploads/2013/03/PS3_Error_Propagation_sp13.pdf

  2. http://www.itl.nist.gov/div898/handbook/mpc/section5/mpc552.htm

  3. http://lben.epfl.ch/files/content/sites/lben/files/users/179705/Error%20Propagation_2012.pdf


Okay, I eventually figured it out by myself: The possible errors of the components have to be propagated separately and then summed up:

model error=slope error+value error+intercept error

This looks like this:

dy = (ds*(x-X0)) + (s*dx) + (s*dX0)

So this would yield for my example:

dy=(0.01*(25-5)) + (0.9*2) + (0.9*1.5)



This is consistent with the value I would obtain when solving the linear model with all possible combinations of values (results range between 16.333 and 19.684).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.