Time-Averaged Drift Approximations are Inconsistent for Inference in Drift Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Sicheng, Fengler, Alexander, Frank, Michael J., Harrison, Matthew T.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914194143576064
author Liu, Sicheng
Fengler, Alexander
Frank, Michael J.
Harrison, Matthew T.
author_facet Liu, Sicheng
Fengler, Alexander
Frank, Michael J.
Harrison, Matthew T.
contents Drift diffusion models (DDMs) have found widespread use in computational neuroscience and other fields. They model evidence accumulation in simple decision tasks as a stochastic process drifting towards a decision barrier. In models where the drift rate is both time-varying within a trial and variable across trials, the high computational cost for accurate likelihood evaluation has led to the common use of a computationally convenient surrogate for parameter inference, the time-averaged drift approximation (TADA). In each trial, the TADA assumes that the time-varying drift rate can be replaced by its temporal average throughout the trial. This approach enables fast parameter inference using analytical likelihood formulas for DDMs with constant drift. In this work, we show that such an estimator is inconsistent: it does not converge to the true drift, posing a risk of biasing scientific conclusions drawn from parameter estimates produced by TADA and similar surrogates. We provide an elementary proof of this inconsistency in what is perhaps the simplest possible setting: a Brownian motion with piecewise constant drift hitting a one-sided upper boundary. Furthermore, we conduct numerical examples with an attentional DDM (aDDM) to show that the use of TADA systematically misestimates the effect of attention in decision making.
format Preprint
id arxiv_https___arxiv_org_abs_2512_10250
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Time-Averaged Drift Approximations are Inconsistent for Inference in Drift Diffusion Models
Liu, Sicheng
Fengler, Alexander
Frank, Michael J.
Harrison, Matthew T.
Methodology
Applications
Computation
Drift diffusion models (DDMs) have found widespread use in computational neuroscience and other fields. They model evidence accumulation in simple decision tasks as a stochastic process drifting towards a decision barrier. In models where the drift rate is both time-varying within a trial and variable across trials, the high computational cost for accurate likelihood evaluation has led to the common use of a computationally convenient surrogate for parameter inference, the time-averaged drift approximation (TADA). In each trial, the TADA assumes that the time-varying drift rate can be replaced by its temporal average throughout the trial. This approach enables fast parameter inference using analytical likelihood formulas for DDMs with constant drift. In this work, we show that such an estimator is inconsistent: it does not converge to the true drift, posing a risk of biasing scientific conclusions drawn from parameter estimates produced by TADA and similar surrogates. We provide an elementary proof of this inconsistency in what is perhaps the simplest possible setting: a Brownian motion with piecewise constant drift hitting a one-sided upper boundary. Furthermore, we conduct numerical examples with an attentional DDM (aDDM) to show that the use of TADA systematically misestimates the effect of attention in decision making.
title Time-Averaged Drift Approximations are Inconsistent for Inference in Drift Diffusion Models
topic Methodology
Applications
Computation
url https://arxiv.org/abs/2512.10250