1
$\begingroup$

I have a quantity defined as: $P_{frac} = \frac{F_{max}-F_{min}}{F_{max}+F_{min}}$

I also have the value for $F_{max}$, $F_{min}$, and their statistical errors.

How can I calculate the error for $P_{frac}$?

$\endgroup$
  • $\begingroup$ What, precisely, do you mean by "their statistical errors"? Are $F_{max}$ and $F_{min}$ dependent? $\endgroup$ – Glen_b -Reinstate Monica Mar 11 '14 at 8:54
  • $\begingroup$ They are results from a measurement, and then they also are reported with their error. They are not dependent. $\endgroup$ – Py-ser Mar 11 '14 at 9:04
  • $\begingroup$ They are not dependent. -- out of curiosity, how do you know this? $\endgroup$ – Glen_b -Reinstate Monica Mar 11 '14 at 9:49
  • $\begingroup$ They are the maximum and the minimum value of a series of data. $\endgroup$ – Py-ser Mar 11 '14 at 11:46
  • $\begingroup$ In which case they're very likely not independent (though if the samples are large, it may be a reasonable approximation). $\endgroup$ – Glen_b -Reinstate Monica Mar 11 '14 at 19:34
1
$\begingroup$

If you can work it out for $R=\frac{F_{max}}{F_{min}}$, and you can work it out for a reciprocal (multiplicative inverse), then rewrite:

$$P_{frac}=\frac{1}{1+\frac{2}{R-1}}$$

Which involves only shifts, inverses and a scalar multiplication.

$\endgroup$
  • $\begingroup$ Thanks Glen, but... what about then? $\endgroup$ – Py-ser Mar 11 '14 at 11:46
  • 1
    $\begingroup$ I don't think I should say more because at present: (i) my very first question in comments remains unanswered; and (ii) you've made some statements that don't seem consistent with each other (and which I don't presently have enough information to choose between). So I don't know for certain what I am supposed to be calculating nor the circumstances under which I am doing it. If you clarify your question sufficiently, I might be able to say more. $\endgroup$ – Glen_b -Reinstate Monica Mar 11 '14 at 19:36
0
$\begingroup$

addition and subtraction -> add absolute errors

multiplication and division -> add relative errors

so, $\delta_{frac}=\frac{\Delta_{max}+\Delta_{min}}{F_{max}-F_{min}}+\frac{\Delta_{max}+\Delta_{min}}{F_{max}+F_{min}}=(\Delta_{max}+\Delta_{min})\frac{2F_{max}}{F_{max}^2-F_{min}^2}$

absolute error $\Delta_{frac}=(\Delta_{max}+\Delta_{min})\frac{2F_{max}}{(F_{max}+F_{min})^2}$

$\endgroup$
  • $\begingroup$ In light of the comments to the question, which point out (among other things) the dependence of the variables and imply strongly asymmetric distributions for them, this approach seems unsuitable. $\endgroup$ – whuber Mar 13 '14 at 13:21
  • $\begingroup$ it was described as a measurement error initially, i wasn't following the comments. i think "my" formula will give a very good approximation in any case $\endgroup$ – Aksakal Mar 13 '14 at 13:25

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.