Extending the diffusion/Wiener model analysis to close the boundary condition over time

I’m working with Henrik Singman’s tutorial on how to use brms and rstan to fit a diffusion/wiener model. What I’d like to do is to manipulate the boundary separation (alpha) so that instead of it being a constant boundary over time, it should have a slope. In my specific case, I want the boundaries to become smaller and smaller so that the subject acts on less and less evidence as time goes by. (I could also simulate the same process by gaining the random-walk process over time.)

Now, I’m not quite sure how one could implement this. Would it be possible to only manipulate the model formula statement or would one have to start digging into the source code of rstan to manipulate the Wiener function itself? That is, is this achievable without too much hassle? I’m having a somewhat hard time to understand where to begin here.

1 Like

You can write your own densities in Stan programs, which you can use in RStan. But if you meant RStanArm rather than RStan, then yes, you’d have to dive into the underlying code.

1 Like

There is no analytical solution for diffusion model whose drift rate or boundart sparation is time varying. Some people used KDE to fit these models.

1 Like

LBA models do have analytic solutions for at least the more common configurations, right?

Hi, I have similar problem. How did you finally solve that?