Take the 2-minute tour ×
Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

I have a series of functions, each one supposedly representing the density of a random variable across agents. Each function also has a domain, which describes what values of the random variable are valid.

Now, if I remember my stats classes correctly, if I take the integral of one of the functions across the values described by the function's domain, I should get a value of 1.0. This does not happen however.

Is there a normalization technique that can turn a function into a true probability density, yet maintains the shape of the function?

All the functions are of the form $\frac{a}{bx}+c$, where $x$ is the random variable, and $a,b,c$ are varying constants.

share|improve this question
add comment

1 Answer

up vote 12 down vote accepted

If you have a non-negative integrable function $f$ with domain $D$ such that

$$ k = \int_{D} f(x) dx < \infty $$

Then $f(x)/k$ is a probability density on $D$. The value $k$ is known as the normalizing constant.

Edit: In your example you said that $f(x) = \frac{a}{bx} + c$ for known constants $a,b,c$. In that case, the indefinite integral is simple to compute and the normalizing constant would be

$$ k = \left[ \frac{a \log(x) }{b} + cx \right]_{D}$$

if $D$ is an interval $(A,B)$ then this simplifies to

$$ k = \frac{a}{b} \cdot \log \left( \frac{B}{A} \right) + c(B-A) $$ Therefore $$ g(x) = \frac{\frac{a}{bx} + c}{\frac{a}{b} \cdot \log \left( \frac{B}{A} \right) + c(B-A)}$$ is a probability density on $(A,B)$.

share|improve this answer
    
Thanks, this is exactly what I was thinking of. –  Graham Jul 20 '11 at 19:07
add comment

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.