psychopy-pest

A PsychoPy plugin that adds PEST (Parameter Estimation by Sequential Testing; Taylor & Creelman, 1967) as psychopy.data.PestHandler, alongside PsychoPy's own StairHandler and QuestHandler.

Taylor, M. M., & Creelman, C. D. (1967). PEST: Efficient estimates on probability functions. The Journal of the Acoustical Society of America, 41(4A), 782–787. https://doi.org/10.1121/1.1910407

It implements the same procedure as the MATLAB toolbox in this repository and gives identical results: the Wald sequential test (or an optional N-down/M-up rule), the five step-size rules, and termination when a reversal would halve the step below the minimum. See the main README for the rules.

Installation

In the Python environment that runs PsychoPy:

pip install "psychopy-pest @ git+https://github.com/ZhenxianLi/PEST_Toolbox#subdirectory=python"

In the PsychoPy app, open Tools → Plugins/packages manager → Packages and install the same address, or run the command above in the app's terminal.

Usage with PsychoPy

from psychopy import plugins
plugins.scanPlugins()             # find newly installed plugins
plugins.loadPlugin('psychopy-pest')
from psychopy import data

staircase = data.PestHandler(
    startVal=-20,         # intensity of the first trial
    startStepSize=8,
    minStepSize=1,        # run ends when a reversal halves the step below this
    maxStepSize=16,
    pThreshold=0.75,      # Wald test target
    waldLimit=1)

for intensity in staircase:          # stops when the PEST run ends
    isCorrect = runTrial(intensity)  # your trial code
    staircase.addResponse(isCorrect)

print(staircase.threshold)
staircase.saveAsExcel('pest_run.xlsx')

PestHandler is a subclass of StairHandler, so addOtherData, the saveAs* methods and ExperimentHandler.addLoop work as usual. A correct response steps the intensity down, and steps are linear: express the intensity in the unit PEST should step in (for example, dB).

In Builder, enable the plugin in the plugin manager and create the handler in a Code component.

Usage without PsychoPy

The algorithm itself has no dependencies:

from psychopy_pest import PEST

pest = PEST(initial_level=-20, initial_step_size=8, min_step_size=1)
while not pest.finished:
    pest.update(run_trial(pest.level))
print(pest.threshold, pest.history)

Development

pip install -e ".[test]"
pytest

The handler tests run when PsychoPy is installed and are skipped otherwise. examples/simulated_observer.py runs both classes against a simulated observer.

License

Non-commercial research and educational use only; see LICENSE.