Cumulative distribution function is pretty much known and well covered from undergrad to grad probability texts. Specifically if $\mathbf P$ is a probability measure on $(\mathbb R, \mathfrak B_\mathbb R),$ then cdf is defined as

$$\mathsf F(x):= \mathbf P((-\infty, x])\tag 1\label 1.$$

However, early texts like those by Kolmogorov, Uspensky, Gnedenko defined it (borrowing from Kolmogorov's notations) as

$$\mathsf F^{(x)}(a) := \mathbf P(-\infty, a).\tag 2\label 2$$

From $\eqref 1,~~\forall~a<b,$

$$\mathbf P(\{\omega:X(\omega)\in(a, b]\}) = \mathsf F(b)-\mathsf F(a).\tag 3\label 3$$

While from $\eqref 2,$

$$\mathbf P\{x\in[a,b)\} = \mathsf F^{(x)}(b)-\mathsf F^{(x)}(a).\tag 4\label 4$$

While $\eqref 1$ is right-continuous, $\eqref 2$ is left-continuous.

In general too, while I was acquainted with Stieltjes premeasure on Borel field $\mathfrak B_{(~~~]}(\mathbb R)$ which is generally defined for a non-decreasing, right continuous $F:\mathbb R\to\mathbb R$ as

$$\chi_F(a, b] := F(b)- F(a),\tag 5\label 5$$

in Schilling's measure theory text, the author defines Stieltjes function for monotonically increasing and left-continuous $F$ (he calls it Stieltjes function) as

$$\nu_F([a, b)):= F(b)-F(a).\tag 6\label 6$$

My question is why Kolmogorov and his other contemporaries defined the cdf as $\eqref 2$ and not as $\eqref 1$ and how $\eqref 1$ became the more common face as we are accustomed to. Which author first used $\eqref 1$ as the definition rather than $\eqref 2?$ Was Kolmogorov the first one in history to define cdf as $\eqref 2?$ Did Borel, Jordan introduce any formal definition on the same prior to Kolmogorov in their treatises?

In a nutshell, my query revolves around the timeline of the introduction of cdf and its development to the one we are familiar with today.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.