1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Error propagation log2

  1. Jun 15, 2015 #1
    Hi there,

    I have a quick question to report some numbers on an experiment. I made measurements of fluorescence in a titration of a chemical. The titrations were 1:2 serial dilutions so I report each fluorescence as a function of the log2 concentration:

    concentration chemical x: 1 , 0.5, 0.25
    reported concentration chemical x (log2): 0, -1,-2
    fluorescence measurement: 5, 10, 25

    I’m interested in reporting the concentrations of a chemical that lead to a specific fluorescence. So, let’s say I want to report when the fluorescence reaches 10, in this case that would be -1.

    So keeping that in mind: When I make error propagation calculations should I use the log2 or the linear measurements?

    Sometimes I need to interpolate to get the reported concentrations. For instance, when I want to get the concentration that leads to fluorescence 15. Does that change anything for the calculations?
  2. jcsd
  3. Jun 15, 2015 #2


    User Avatar
    Education Advisor
    Gold Member

    Welcome to the forum.

    It depends on context.

    Generically, you should report what you measured, and the error in your measurement. If you measured x then you should report x and the error in x. If you measured log2 of x, then you should report log2 of x and the error in log2 of x.

    If you are calculating something based on the concentration, then if you are using x in the calculation you should use the uncertainty in x. If you are using log2 of x in the calculation, then you should use the error in log2 of x.

    Does this help?
  4. Jun 15, 2015 #3
    Thanks for your reply. So if I understand correctly, as long as I don't mix logs and linear measurements I should be all set for what I'm reporting.

    I'm still a confused though when I want to compute different metrics, like the error propagation or the percent difference ( i.e. , difference/average)

    So for instance if I have the following replicate measurements and I want to compute the percent difference:

    Replicate 1: 0.0625
    Replicate 2: 0.0725

    Replicate 1 in log2: -4
    Replicate 2 in log2: -3.7859

    Percent difference in linear: (0.0725-0.0625) / ( (0.0725+0.0625)/2)=

    Percent difference in log2: abs(-4-(-3.7859))/((-4+-3.7859)/2)

    Why are these two numbers different when I'm reporting a percentage? Shouldn't this percentage be the same since I'm only changing the base but not the values of the measurements?

    Thanks again!
  5. Jun 15, 2015 #4


    User Avatar
    Science Advisor
    Gold Member

    Non-constant functions do not, in general, preserve ratios, e.g., : ## \sqrt {\frac {81}{4}}=4.5 \neq \frac {81}{4}=20.25 ##. I think this has to see with the derivative of these functions not being constant, i.e., only when the function ##f ##, as below, is linear.

    Basically, few functions preserve ratios, i.e., few functions satisfy:

    ## \frac {a}{b}=\frac {f(a)}{f(b)} ####
    Last edited: Jun 15, 2015
  6. Jun 16, 2015 #5
    Oh yes, this makes a lot of sense. Thank you WWGD
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook