Exercises 4
Exercise 4.1
Consider the Laplace family of distribution, \(\mathsf{Laplace}(\nu, \tau)\), with density \[\begin{align*} g(x; \nu, \tau) = \frac{1}{2\tau} \exp\left(- \frac{|x-\nu|}{\tau}\right), \qquad \nu \in \mathbb{R}, \tau > 0 \end{align*}\] as a candidate distribution for rejection sampling from \(\mathsf{Gauss}(0,1).\)
- Provide an inversion sampling algorithm to generate from \(\mathsf{Laplace}(\nu, \tau).\)
- Can you use the proposal to generate from a standard Gaussian? for Student-\(t\) with 1 degree of freedom? Justify your answer.
- Consider as proposal a location-scale version of the Student-t with \(3\) degrees of freedom. Find the optimal location and scale parameters and the upper bound \(C\) for your choice.
- Use the accept-reject to simulate 1000 independent observations and compute the empirical acceptance rate.