Suppose we have a uniform random variable $U$ which is defined on the $[0,1]$ interval. Consider the transformation:$$X=k\times \log(U)$$

How would I go about calculating the interval on which $X$ is defined?

I guess on one hand, the upper limit would be given as $k\times \log(1)=0$ . However, I cannot evaluate $k\times \log(0)$. Any ideas? Thanks!


The Intervall should $(-\infty,0]$, Since the right limit of log at zero is $-\infty$

Edited due to dsaxton's comment: Note that the probability that U takes any value $c \in [0,1]$ is 0, or in math notation $P(U=c)=0$ for every $c \in [0,1]$

Why do we take limits, (do we?)? First, let me remark that log is bijective strictly increasing function from $(0,1] \to (-\infty,0]$. Using some math one could define the continuous extension of in zero log to be $log(0):= -\infty$ However, in that case, $P(X=-\infty)=P(U=0)=0$, and the value $-\infty$ won't play a role.

However, to determine the image of $X$, we employ the range of log on $(0,1]$ being $(-\infty,0]$. In both intervals, and in the determination of the image of log on $(0,1]$ limits are involved

| cite | improve this answer | |
  • $\begingroup$ Im not sure I follow..U does take on a value of 0 with probability density =1. This is not the case of the transformation of U $\endgroup$ – ChinG Feb 23 '16 at 16:02
  • $\begingroup$ Thanks for your response, but why do we evaluate the limit of the function at the value instead of the function at the value? $\endgroup$ – ChinG Feb 23 '16 at 16:11
  • 2
    $\begingroup$ @ChinG The density of $U$ is defined at one (entirely arbitrarily it turns out) but $P(U = 0) = 0$ so you never have to worry about $\log(0)$. $\endgroup$ – dsaxton Feb 23 '16 at 16:12
  • $\begingroup$ Thanks a lot for the lucid explanation. One final question: why are we involving limits of the function at the value here instead of evaluating the function at the value itself? $\endgroup$ – ChinG Feb 23 '16 at 16:15

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.