Modeling Behavior in Truth Value Judgment Task Experiments

Brandon Waldon and Judith Degen

Requirements:

• WebPPL (https://github.com/probmods/webppl) (to perform Bayesian Data Analysis)
• R/RStudio + relevant packages (to analyze and visualize BDA results)

Run BDA to get posteriors over parameters:

Binary condition

webppl inference_scripts/binary_script.wppl --random-seed 1571250871 --require webppl-json --require './node_modules/mathjs'

Ternary condition

webppl inference_scripts/tertiary_script.wppl --random-seed 1571250622 --require webppl-json --require './node_modules/mathjs'

Quaternary condition

webppl inference_scripts/quaternary_script.wppl --random-seed 1571251139 --require webppl-json --require './node_modules/mathjs'

Quinary condition

webppl inference_scripts/quinary_script.wppl --random-seed 1571251077 --require webppl-json --require './node_modules/mathjs'

Generate predictive distributions using MAP estimates of parameters:

First, open analysis/analysis.r to generate the MAP estimates as .json objects (stored in analysis/inference_results). Then run:

webppl inference_scripts/binary_forward.wppl --random-seed 123 --require webppl-json
webppl inference_scripts/ternary_forward.wppl --random-seed 123 --require webppl-json
webppl inference_scripts/quaternary_forward.wppl --random-seed 123 --require webppl-json
webppl inference_scripts/quinary_forward.wppl --random-seed 123 --require webppl-json