two warnings from fitting the drift diffusion model

[Daniel Reznik]

Dear group,

I am fitting a drift diffusion model and I get two following warnings from Stan sampler in the initialization phase (both go away after a few warmup iterations):

1 - wiener_lpdf: Boundary separation is inf, but must be finite!

2 - wiener_lpdf: Random variable  = 0.327, but must be greater than nondecision time = 0.327. (other values appear as well.)

Is there a way to fix it? Is it kosher to ignore these? At the end, I get reasonable estimates after sampling, but these warning make me doubt my results.

Thank you so much!

[Haines, Nathaniel B.]

Hi Daniel,

It is hard to say exactly what is going on, but generally these messages come up when the initial values are in a part of the parameter space that leads to something like NaN or Inf values. That said, once the sampler is out of the warm-up stage, so long as your convergence diagnostics are looking good, you should be fine! 

As far as a "fix", you may be able to set different initial parameter values to avoid these messages altogether, but that is not necessary. 

Let me know if that answers your question!

Best,

Nate

---

From: hbayesd...@googlegroups.com <hbayesd...@googlegroups.com> on behalf of Daniel Reznik <rezn...@gmail.com>
Sent: Tuesday, June 30, 2020 6:32 PM
To: hbayesdm-users <hbayesd...@googlegroups.com>
Subject: two warnings from fitting the drift diffusion model

 

--
You received this message because you are subscribed to the Google Groups "hbayesdm-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbayesdm-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hbayesdm-users/a6ca8774-7155-45d6-aec9-6f3ebcfa0140o%40googlegroups.com.

[Daniel]

Thank you! It's challenging to use different initial values, since I am fitting a few models at the same time ( and I'm note sure if I can initiate each one of then at a different value).

I'll be more careful in examining convergence diagnostics!

On Thursday, July 2, 2020 at 5:37:59 PM UTC+3, Haines, Nathaniel B. wrote:

Hi Daniel,

It is hard to say exactly what is going on, but generally these messages come up when the initial values are in a part of the parameter space that leads to something like NaN or Inf values. That said, once the sampler is out of the warm-up stage, so long as your convergence diagnostics are looking good, you should be fine! 

As far as a "fix", you may be able to set different initial parameter values to avoid these messages altogether, but that is not necessary. 

Let me know if that answers your question!

Best,

Nate

---

From: hbayesd...@googlegroups.com <hbayesd...@googlegroups.com> on behalf of Daniel Reznik <rezn...@gmail.com>
Sent: Tuesday, June 30, 2020 6:32 PM
To: hbayesdm-users <hbayesd...@googlegroups.com>
Subject: two warnings from fitting the drift diffusion model

 

Dear group,

I am fitting a drift diffusion model and I get two following warnings from Stan sampler in the initialization phase (both go away after a few warmup iterations):

1 - wiener_lpdf: Boundary separation is inf, but must be finite!

2 - wiener_lpdf: Random variable  = 0.327, but must be greater than nondecision time = 0.327. (other values appear as well.)

Is there a way to fix it? Is it kosher to ignore these? At the end, I get reasonable estimates after sampling, but these warning make me doubt my results.

Thank you so much!

--
You received this message because you are subscribed to the Google Groups "hbayesdm-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to hbayesd...@googlegroups.com.