Negative linear collapsing boundary angle - Interpretation?

[jainraj]

I fit a linear collapsing boundary model to experimental data in which participants have to respond within a maximum time (which they are aware of!).

The estimated theta is coming to be close to -0.1 (lower bound for the LANs). I am wondering how to interpret this result. I was expecting some positive value. Should I try a non-collapsing version, then?

PS:

1. I am not using HSSM/HDDM - I use the LAN directly. (The question still stands - how to interpret negative angle).
2. LAN - Race model with 4 choices, same starting point across all choices with linear collapsing boundary.

[frankmj]

Hi, probably this shouldn;'t be an HSSM discussion as there are many reasons one might obtain some fits to a model depending on the data and the model set up and the inference method so it is pretty hard to diagnose. That said: - the usual strategies are to check that you can recover ground truth parameters from simulated data with your method, and that when you fit the data you can reproduce core features of it (RT quanitles, error rates) with your fitted parameters - note that race models give you a kind of urgency for free without requiring collapsing bounds - if all accumulators have some constant intercept on drift then these will continue to rise throughout the course of a trial toward the boundary independent of the evidence - so that any impact of noise will be greater with more time, similar to collapsing bounds.. in contrast when using a DDM the drift rate accumulates the difference in evidence between options and so does not have this same property unless you had a collapsing bound. So when you have a race model you might not need a collapsing bound to account for the data (and hitting the negative boundary might be trying to compensate for the effect I described above).

…

On Thu, Aug 15, 2024 at 5:54 AM Raj V Jain ***@***.***> wrote: I fit a linear collapsing boundary model to experimental data in which participants have to respond within a maximum time (which they are aware of!). The estimated theta is coming to be close to -0.1 (lower bound for the LANs). I am wondering how to interpret this result. I was expecting some positive value. Should I try a non-collapsing version, then? PS: 1. I am not using HSSM/HDDM - I use the LAN directly. (The question still stands - how to interpret negative angle). 2. LAN - Race model with 4 choices, same starting point across all choices with linear collapsing boundary. — Reply to this email directly, view it on GitHub <#549>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG7TFEP6CFUK5U65L54X5TZRR3E7AVCNFSM6AAAAABMR5IPL6VHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZXGA2TKNZWHA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[jainraj]

Thank you, Professor, for the suggestion and the interpretation.