Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have a prediction $f(x)$ of some continuous process variable, based on an input variable $x$ (think: location). The prediction is incorrect, with the error being normal distributed with expected value $\mu$ and standard deviation $\sigma$.

Hence, the probability density function of $f(x)$ should be

$$\frac{1}{\sigma\sqrt{2\pi}}e^{-\frac{((f(x)-\mu)/\sigma)^2}{2}}$$

Is this correct? (No it is not, see answer below)

Now, I have a measurement $m$ of the process variable, based on an unknown input variable $x_m$. $m$ is assumed to be correct, but quantized to integral numbers.

Given a set of $x_i$ together with their predictions $f(x_i)$, how can I compute a probability that $x_m$ was in the vicinity of $x_i$?

I apologize if the question makes no sense. Comments that help me improve the question are appreciated!

share|improve this question

1 Answer 1

Your proposed density function is not correct. Rather, the probability density function of the error $\varepsilon$ is $$ w\mapsto\frac{1}{\sigma\sqrt{2\pi}} e^{\frac{-((w-\mu)/\sigma)^2}{2}}. $$ If the prediction is $f(x)$, and the true value of the thing to be predicted is $g(x)$, then $$ f(x) = g(x)+\varepsilon. $$ You will never have an exact match since the probability distribution of the error has no discrete component---it's a continuous distribution. So the probability of an exact match is $0$. But you can speak of the probability that the prediction differs by no more than a specified amount from the true value. That probability would depend on $\mu$ and $\sigma$.

share|improve this answer
    
Thanks for the answer! I've tried to improve the question, although I haven't "fixed" that part about the PDF yet. –  stmoebius Aug 1 '12 at 16:16

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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