0
$\begingroup$

I am struggling with a process like this: $$X_t=\begin{cases} \frac{\alpha\omega_t}{\alpha\omega_t+\beta(1-\omega_t)} & \text{with prob } p\\ \frac{(1-\alpha)\omega_t}{(1-\alpha)\omega_t+(1-\beta)(1-\omega_t)} & \text{with prob } 1-p \end{cases}$$ where $\alpha\in(0,0.5)$ and $\beta\in(0.5,1)$. Now introduce the process (which is a Bayesian update) $$ \omega_{t+1}=k+hX_t $$ where $h,k\in[0,1]$ Thus, $X_t$ is a simple Bernoulli trial, but it assumes different values at any time step $t$, since it depends on $\omega_t$. My question is then: Is it possible to compute the distribution of $\omega$? I made some numerical simulations: distribution is not always the same and it changes depending on parameters $\alpha, \beta, h, k$. Following I attach two simulations-example of the density of $\omega$ with different combinations of parameters (sorry if I am rude and i didn't rename axis label and title :D).

Many thanks enter image description here

enter image description here

Improve this question
$\endgroup$
4
  • 1
    $\begingroup$ This question might better fit in the stats forum stats.stackexchange.com. $\endgroup$ Commented May 17, 2023 at 22:08
  • $\begingroup$ A close relative to the density of $\omega_t$ is finding the limiting/stationary distribution as $t$ grows. If the coin flips are iid, the process $\omega_t$ is Markov since it only depends on the previous step. So there is some hope for finding transition density Matrix (if we restrict to discrete times) and then stationary distribution. $\endgroup$ Commented May 17, 2023 at 22:21
  • $\begingroup$ this is quite a complicated definition also because if the jump step k is too large, then $\omega_t$ quickly heads to $\frac{\beta}{\beta-\alpha}$ and thus the Xt diverges. This boundary effect makes it a bit hard to study. $\endgroup$ Commented May 18, 2023 at 0:46
  • $\begingroup$ Sorry I forgot to mention that $\omega$ lies between 0 and 1, so X cannot diverge. $\endgroup$ Commented May 19, 2023 at 20:37

0

Reset to default

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.