5
$\begingroup$

I have this problem: In betting situations one is often interested in odds, referring in the thumbtack tossing $\theta / (1 - \theta)$. Alternatively one may consider the log-odds:

$\lambda=log(\theta/(1-\theta))$

Show that a uniform distribution for $\lambda$ implies the following distribution for $\theta$:

$p(\theta)=\theta^{-1}(1-\theta)^{-1}$

What problems are associated with this distribution if it is used as a prior in the thumbtack tossing problem?

Now I have the thumbtack thingy, I know what it will happen if we use that p as prior, but the uniform for $\lambda$ implies the p($\theta$) I don't get it, I have tried to get to the p($\theta$) from both sides but I always get to $log(\theta)-log(1-\theta)$ and can't get out, any help??

$\endgroup$
6
$\begingroup$

$$\frac{d\lambda}{d\theta}=\frac{1}{\theta} + \frac{1}{1-\theta} = \frac{1}{\theta(1-\theta)}$$ $$f(\lambda)\,d\lambda\propto d\lambda \quad \text{(improper uniform; aka Lebesgue measure)}$$ $$g(\theta)\,d\theta =f(\lambda)\,d\lambda=f(\lambda(\theta))\frac{d\lambda}{d\theta}\,d\theta\propto \frac{1}{\theta(1-\theta)}d\theta$$ $$g(\theta)\propto\frac{1}{\theta(1-\theta)} \quad \text{(improper Haldane's prior; see also Jaynes)}$$

| cite | improve this answer | |
$\endgroup$
  • 2
    $\begingroup$ Zen's cryptic replies are invitations to do the work myself and are fully half the reason I read C.V. $\endgroup$ – Sycorax Mar 14 '14 at 17:55
  • $\begingroup$ I wonder if this is a good or a bad thing... $\endgroup$ – Zen Apr 12 '14 at 20:26
  • $\begingroup$ @Zen You're one of my favorite contributors. $\endgroup$ – Sycorax Sep 6 '14 at 18:30

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.