I am looking at the CIR++ model which is described in interest rate models by Brigo et al, and was wondering on how to actually simulate from this model. The model reads


where $x$ follows a CIR model.

To keep it as simple as possible i thought i would simulate the x process according to a 'truncated' Euler scheme, i.e.

$$x_{t_i}=x_{t_{i-1}}+\kappa(\theta-x_{t_{i-1}})\Delta t+\sigma \sqrt{x_{t_{i-1}}^+} \Delta t Z_t,$$

where $Z_t$ is $N(0,1)$. My question then is, would it make sense do to so and simply add $\phi(t)$ accordingly, meaning


  • $\begingroup$ What is $\phi(t)$ ? $\endgroup$ – Richard Mar 30 '16 at 11:44
  • $\begingroup$ why not $r_{t_i}=x_{t_i}+\phi(t_i)$? $\endgroup$ – Gordon Mar 30 '16 at 12:58
  • $\begingroup$ $\phi(t)$ is a determinstic shift function, which is chosen in order to match the initial term structure. You are right Gordon, that is a typo. $\endgroup$ – Bohlke Mar 30 '16 at 13:12
  • $\begingroup$ Sorry but I find your notations a bit sloppy. What does the first term in the rhs of your Euler discretisation represent? It should be $x_{t_i} $. Plus, this is not exactly the full truncation method, $x_{t_i} $ involved in the mean reversion component should be truncated to $x_{t_i}^+ $. The rest seems ok but could you clear things up / fix your notations to be sure? $\endgroup$ – Quantuple Mar 30 '16 at 16:14
  • $\begingroup$ Sorry, things went a little fast. But the first term should be $x_{t_{i-1}}$. The schemes that I have seen in the literature normally only truncates the component inside the square-root, to avoid taking the square-root of negative numbers. But I see your point, a full truncation could be done by following your method. What I was actually doubting was whether the last equation makes sense, where I add the $\phi$ function $\endgroup$ – Bohlke Mar 30 '16 at 17:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.