# Characterization of Probability Density Function

Jump to navigation
Jump to search

## Theorem

Let $\struct {\Omega, \Sigma, \Pr}$ be a probability space.

Let $X: \Omega \to \R$ be a continuous random variable on $\struct {\Omega, \Sigma, \Pr}$.

Let $\Omega_X = \Img X$, the image of $X$.

Let the **probability density function** of $X$ is the mapping $f_X: \R \to \closedint 0 1$ be defined as:

- $\forall x \in \R: \map {f_X} x = \begin {cases} \displaystyle \lim_{\epsilon \mathop \to 0^+} \frac {\map \Pr {x - \frac \epsilon 2 \le X \le x + \frac \epsilon 2} } \epsilon & : x \in \Omega_X \\ 0 & : x \notin \Omega_X \end {cases}$

Suppose that the cumulative distribution function of $X$ defines a continuously differentiable real function $F_X: x \mapsto \map \Pr {X \le x}$.

Then the **probability density function** of $X$ is the mapping $f_X: \R \to \closedint 0 1$ satisfies:

- $\dfrac{\d}{\d x} \map{F_X}{x} = \map {f_X} x$.

## Proof

\(\displaystyle \map \Pr {x-\frac{\epsilon} 2 \le X \le x + \frac {\epsilon} 2 }\) | \(=\) | \(\displaystyle \map{F_X}{x + \frac{\epsilon} 2} - \map{F_X}{x - \frac{\epsilon} 2} + \map \Pr {X= x- \frac {\epsilon} 2 }\) | |||||||||||

\(\displaystyle \) | \(=\) | \(\displaystyle \map{F_X}{x + \frac{\epsilon} 2} - \map{F_X}{x - \frac{\epsilon} 2}\) | Probability of Continuous Random Variable at Single Point is Zero | ||||||||||

\(\displaystyle \leadsto \ \ \) | \(\displaystyle \map {f_X}{x}\) | \(=\) | \(\displaystyle \lim_{\epsilon \to 0^+} \frac{\map {F_x}{x+\frac \epsilon 2} - \map {F_x}{x - \frac \epsilon 2} } {\epsilon}\) | ||||||||||

\(\displaystyle \) | \(=\) | \(\displaystyle \dfrac{\d}{\d x} \map{F_X}{x}\) | Continuous Derivative as Average of One-Sided Derivatives |

$\blacksquare$

## Sources

- 2001: Michael A. Bean:
*Probability: The Science of Uncertainty*: $\S 4.1$