2
$\begingroup$

http://imgur.com/FeFf3e9

The imgur link is to a screenshot of the relevant section in my text. I have trouble understanding how if $H(x, \infty)=F(x)$ is the marginal distribution of $x$, how $F(x) = x, 0 < x < 1$ is the uniform marginal. How can the marginal distribution be uniform if it equals $F(x)=x$? Furthermore how can a CDF function be uniform?

What does "uniform" refer to in this text?

$\endgroup$
  • 1
    $\begingroup$ If at all possible, I would recommend simply putting the relevant text from your image directly in your question, and cite the source. $\endgroup$ – Patrick Coulombe Mar 5 '14 at 0:05
2
$\begingroup$

Uniform margins are true of every copula, not just FGM copulas (not sure why that book leaves Gumbel out of the usual copula name).

In this case, uniform means "Has a uniform distribution" - that is, the density is constant over some interval, and in the case of copulas, where that interval is $[0,1]$, the corresponding distribution function is of the form $F_X(x)=x$ over that interval (and is 0 to the left of it and 1 to the right of it). Note that the cdf, $F_X(x)=x$ corresponds to a constant density, $f$, which is why the distribution is called 'uniform'.

Note that copulas have uniform $[0,1]$ marginals by definition. The particular copula you refer to has been chosen to fit with the definition. That it does so is easily seen; substitute $y=1$ into $H$ to see it for $X$, and $x=1$ into $H$ to see it for $Y$. Alternatively you can compute the marginals from the joint density by integration.

A search here for copulas will get you a lot more basic information relating to copulas.

$\endgroup$
  • $\begingroup$ Thanks for your answer, but I'm confused if the marginals have a uniform distribution, how can Fx(x)=x? How is that uniform, wouldn't the distribution depend on x and not just be constant? $\endgroup$ – sayyid monroe Mar 5 '14 at 0:12
  • $\begingroup$ I don't understand your question. ... unless you're confused about the difference between density and distribution perhaps. In any case, I am really not sure what you're asking. The density $f$, for $F(x)=x$ is constant (i.e. uniform). Did you read the link? $\endgroup$ – Glen_b Mar 5 '14 at 0:37
  • $\begingroup$ I guess my question does concern the difference between the density and distribution. I calculated the density by taking the integral of the joint density and found that it is indeed 1 and thus uniform. However I don't understand the statement that the marginal distributions are H(x,∞) = F(x). F(x)=x and that is not uniform. $\endgroup$ – sayyid monroe Mar 5 '14 at 1:19
  • $\begingroup$ integrate the marginal density $(\,f_X(x)=1,\quad 0\leq x\leq 1;\,0\text{ elsewhere })$, and you obtain the marginal distribution function, $F_X(x)=x, \quad 0\leq x\leq 1;\,=0,\quad x<0;\,=1,\quad x>1$, or equivalently, differentiate $F$ to get $f$. If you're not familiar with these basic relationships, you should not be trying to understand copulas yet, because you don't have the necessary tools. $\endgroup$ – Glen_b Mar 5 '14 at 1:29
  • $\begingroup$ Ok. Thanks for walking me through it all; that was actually very simple. The fact that FX(x)=x=F(x) really messed me up as I kept getting hung up on the fact that F(x) is a CDF of X and didn't understand how that equaled the marginal distribution of X. $\endgroup$ – sayyid monroe Mar 5 '14 at 1:35

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.