I don't quite understand what people mean by let "$X$ be a random variable WITH a cdf/pdf". For example, there is a question that says: "Let X be a random variable with the 3-parameter Weibull pdf and cdf"

Suppose I say: Let $X$ be a random variable with Gaussian CDF.

What does that mean exactly? $X$ is a function that maps from the event space to a real number. What does it mean to "connect" it, "link" it, "equip" it, WITH a CDF? It is already a function, what does it mean to let it hook up with another function, why don't we just deal with $X$ directly.

Then I look at the CDF:

$f_X(x) = \int_A \exp(-x^2/2)dx$

I ask myself: Is "$\exp(-x^2/2)$" part the random variable? No. Is "$\int_A \exp(-x^2/2)dx$" the random variable? No. Is little $x$ is the random variable? No. What is the difference if I wrote $f(x)$ instead of $f_X(x)$? Nothing happens.

What does the CDF/PDF have to do with the random variable exactly? How do you know which CDF/PDF a random variable "has"?

  • 2
    $\begingroup$ "MSE is a dating site" is quite an interesting username $\endgroup$ – user223391 Mar 17 '16 at 5:21
  • $\begingroup$ "It is already a function...why don't we just deal with X directly." That function only tells you about possible outcomes, but nothing about probabillities - i.e. about how likely each outcome of $X$ is. For that you need a PDF. $\endgroup$ – Winther Mar 17 '16 at 5:36
  • $\begingroup$ The main take-away is that the underlying $\Omega$ (practically) doesn't matter and all you care about is CDF $F_X$ and measure $dF_X$ induced by it. $\endgroup$ – A.S. Mar 17 '16 at 13:47

Yes $X$ is a function from the event space $\Omega$ to the reals $\mathbb{R}$, but there is a bit more: $\Omega$ is a probability space, that is, it has a $\sigma$-algebra and a probability measure. If these terms are unfamiliar to you, loosely speaking this is just saying that every event has a probability associated with it [that satisfies some reasonable properties, e.g., probability of union of two disjoint events is the sum of their respective probabilities].

Note that this induces a probability measure on $\mathbb{R}$ as well. Given any interval $(-\infty,t]$ (or more generally any open set), we can define its probability as the probability of the preimage $X^{-1}((-\infty,t])$, that is, the probability of the subset of $\Omega$ that $X$ maps to $(-\infty,t]$. We often denote this simply as $\mathbb{P}(X \le t)$.

Without this notion of probability, all you would have is a map; you wouldn't have any notion of questions like "what is the probability that $X$ lies in the interval $[0,3]$?" More concretely, if $X$ is the outcome of a roll of a die, then all you have is a map to $\{1,\ldots,6\}$ and no notion of how likely each outcome is. The randomness comes from the underlying probability measure.

Now, if one says "$X$ has CDF $F$," then this is just specifying the underlying probability measure, i.e. it tells you $F(t) := \mathbb{P}(X \le t)$ for any $t$. It turns out that if we know this information, we can find the probability of any Borel set.

Regarding your last few questions ("What is the CDF/PDF of a random variable?") here are the definitions explicitly:

The CDF of a random variable $X$ is a function $F:\mathbb{R} \to [0,1]$ defined by $F(t) := \mathbb{P}(X \le t)$.

Not every random variable has a PDF, but if its CDF is differentiable, then its PDF is defined to be $f(x) := \frac{d}{dt}F(t)$. By calculus, the PDF satisfies $F(t)=\int_{-\infty}^t f(x) \mathop{dx}$.


A random variable formally is a measurable map from a probability space $(\Omega, \sigma, P)$ to $(\Bbb{R}, B(\Bbb{R}), \mu)$ where $\mu$ is Lebesgue measure and $B(\Bbb{R})$ is the borel $\sigma$-algebra.

If $P \ll \mu$ (read $P$ is absolutely continuous wrt $\mu$. Meaning $\mu(S)=0\implies P(S)=0$) and $\Omega$ is $\sigma$-finite, then the Radon Nikodym derivative, $f$ exists and is called the probability density function.

If you want a little more elementary explanation. By "measurable map" we mean that $P(X\leq x)$ is always defined. We simply define the new function $F(x)=P(X\leq x)$ and if the derivative of $F$ exists, then $f$ is the PDF.

  • $\begingroup$ Possibly your symbols are too obscure for the querants level of comprehension. Do you expect MSE is a dating site to understand what $P\ll\mu$ means? $\endgroup$ – Graham Kemp Mar 17 '16 at 5:44

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.