If a sequence of non-negative random variables $X_1, X_2, \dots$ converges almost surely to a random variable $X$, that is $X_n \xrightarrow{a.s} X$ or equivalently $P(\lim\limits_{n\to\infty}X_n=X)=1$, then what can we say about convergence of $1/X_n$?

I believe since $P(\lim\limits_{n\to\infty}1/X_n=1/X) = P(\lim\limits_{n\to\infty}X_n=X)=1$, we conlude $1/X_n \xrightarrow{a.s}1/X$. Is this complete?

  • $\begingroup$ X_n^{-1}=1/X_n? Or what? $\endgroup$ – Martín Vacas Vignolo May 16 '16 at 0:23
  • $\begingroup$ @vvnitram yes, I also edited it. $\endgroup$ – Susan_Math123 May 16 '16 at 0:25
  • 1
    $\begingroup$ is necessary that P(X_n=0)=0, and the same with X... But if this condition is satisfying, the proof is right $\endgroup$ – Martín Vacas Vignolo May 16 '16 at 0:27

Since $X_n \to X$ a.s, there's a set $A$, $P(A)=1$ such that for every $\omega\in A$, $X_n(\omega) \to X(\omega)$. Therefore (calculus) on $A \cap \{\omega:X(\omega)\ne 0\}$ we have $\frac{1}{X_n(\omega)}\to \frac{1}{X(\omega)}$.


  1. Special case. If $X\ne 0$ a.s., then indeed $\frac{1}{X_n}\to \frac{1}{X}$ a.s.
  2. General. $\frac{1}{X_n} \to \frac{1}{X}$ a.s. on $\{\omega:X(\omega)\ne 0\}$ (that is for $\omega$ in that set, possibly except for a subset of measure zero, the limit holds). Note however, that without assuming anything $\{\omega:X(\omega)\ne 0\}$ may be itself of measure zero in which case, the statement is vacuous.

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.