1
$\begingroup$

From pg. 39 of All of Statistics:

If $X_1, \ldots, X_n$ are independent and each has the same marginal distribution CDF $F$, we say that $X_1, \ldots, X_n$ are IID (independent and identically distributed) and we write

$$ X_1, \ldots, X_n \sim F $$

If $F$ has density $f$ we also write $X_1, \ldots X_n \sim f$.

Question 1: Why is the primary focus here on $F$ (the CDF) and not $f$ (the density function)?

Question 2: Is it not already the case that $X_1, \ldots, X_n \sim F$ iff $X_1, \ldots, X_n \sim f$? That is, is it even possible that a collection of random variables $X_i$ could have the same CDFs $F$ but NOT have the same probability density functions $f$ (in either the discrete or the continuous cases)?

$\endgroup$
  • 1
    $\begingroup$ Many random variables do not have a density function. Also, even in the continuous distribution case, the cdf does not determine the density (though admittedly it almost determines the density). $\endgroup$ – André Nicolas Jun 5 '16 at 21:31
  • $\begingroup$ Let me introduce you to my good friend, Cantor. $\endgroup$ – Em. Jun 5 '16 at 21:37
  • $\begingroup$ Andre: I made a comment on Robert's post asking if assuming absolute continuity changed things. An $X$ with Cantor distribution as CDF fails to be absolutely continuous, so there might be hope yet! $\endgroup$ – user1770201 Jun 5 '16 at 22:53
2
$\begingroup$

This is somewhat unfortunate notation, using the same ~ for the CDF and the density.

Every random variable has a CDF, but not every random variable has a probability density function. In particular, a discrete random variable has a probability mass function rather than a density. There are also singular continuous distributions, and mixtures of the different types.

$\endgroup$
  • $\begingroup$ Assuming $X$ is absolutely continuous, do we (at least) then have that $X_1, \ldots, X_n \sim F$ iff $X_1, \ldots, X_n \sim f$? $\endgroup$ – user1770201 Jun 5 '16 at 22:52
  • 1
    $\begingroup$ The density, for an absolutely continuous $X$, is uniquely determined up to values on sets of measure $0$; changing values of the density on sets of measure $0$ does not affect the distribution. $\endgroup$ – Robert Israel Jun 6 '16 at 1:09

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.