Dear community, For fitting DDM I use Wiener First Passage Time Distribution and all works fine in terms of chain diagnostics. However, I cannot run predictive checks since the wiener_rng function is not yet implemented in Stan math library. What can be a current alternative? We tried writing our o…

This works, don’t ask me how I got it to (it was messy). functions { real fs_cdf(real t, real a) { if (a < 1) { reject("a must be >= 1, found a = ", a); } return erfc(inv_sqrt(2 * a * t)); } vector make_vars(real mu) { real mu2 = pow(mu, 2); real t_tilde = 0.…

Thanks! This is what we tried :) didn’t work for us.

I just wrote it up as a Stan program and found that it took an unreasonably long time to produce a result. It compiled and ran, so it “worked” but I doubt it’s usable in a real program.

first of all, thank so much with your help with this! We also wrote something that complies alright, but the predictive checks looked bad. We used the very basic modeling approach, so I think something in the rng was funky.

we also compared results from our custom code and R rwiener code, and we got different outputs.

do you have a wiener model that I could look at? There’s this paper Fast and accurate Monte Carlo sampling of first-passage times from Wiener diffusion models | Scientific Reports which seems promising but there’s some nuance with asymmetric boundaries and I’m not familiar with this density. Having …

@nerpa @WardBrian I think I got the wiener_rng to work. Please test and let me know functions { real fs_cdf(real t, real a) { if (a < 1) reject("a must be >= 1, found a = ", a); return erfc(inv_sqrt(2 * a * t)); } vector make_vars(real mu) { real mu2 = pow(mu, 2); …

this is fantastic! thank you so much! We will test it of course and let you know if it worked. :) :)

There’s some parameterization issue that I’m working through. Hold off on using it for now. It’s nearly there though.

I will not ask :) and thank you so so much, @spinkney ! We will test it and let you know. This is a tremendous help!

Alternative to wiener_rng in Stan

General

WardBrian August 4, 2022, 6:24pm 2

Based on this old thread it seems like it should be possible to implement the wiener_rng function using uniform_rng in pure-Stan. It will likely not be particularly fast, due to the use of a kind of rejection sampling

Topic		Replies	Views
Unable to recover mixture between Wiener-diffusion and uniform distribution Modeling fitting-issues , cognitive-science	14	3244	February 17, 2021
Sampling works in rstan but not cmdstan with a wiener mixture model Interfaces fitting-issues	8	1048	February 18, 2021
Random walk in RStan Modeling rstan , specification , r , prior-choice , priors	11	1457	June 12, 2024
Using the wiener_lpdf() function in the cmdstanr package to implement a diffusion model, the construction of a Markov chain was unsuccessful CmdStan	28	1128	January 29, 2026
Custom likelihood function based on Wiener first passage time density accounting for NoGo responses Modeling cognitive-science	6	1082	April 22, 2021

Alternative to wiener_rng in Stan

Related topics