Estimation of time-varying decision thresholds from the choice and reaction times with no assumption on the shape

When a decision is made based on a series of samples of evidence, the threshold for decision may change over time. Here we propose a fast heuristic algorithm that estimates the time-varying thresholds without making an assumption on their shape. The algorithm gives an approximate best estimate of the time-varying thresholds for all time considered when all other parameters are fixed. The algorithm outperforms conventional methods that gradually adjust the threshold estimates during fitting, when fitting time is limited.