Reanalysis of Iowa Gambling Task data

Original paper:
Killgore WD, Lipizzi EL, Kamimori GH, Balkin TJ. Caffeine effects on risky decision making after 75 hours of sleep deprivation. Aviation, space, and environmental medicine. 2007;78(10):957-62.

---

Directory

1. Data files
2. Scripts
3. Note on data simulations

---

Data files

! NOTE: DATA FILES FOR KILLGORE ET AL. (2007) WILL BE MADE AVAILABLE UPON MANUSCRIPT PUBLICATION.

All data files are located under data subfolder

For real data from Killgore et al. (2007), see data\actual\Killgore2007_GUM4

File/folder	Description
...\raw	Folder contains individual participant data files in .txt format.
Killgore2007_GUM4_merged	Merged usable data across all participants
Killgore2007_GUM4_with_parameters.csv	Merged usable data with computational modelling estimates appended

For simulated data using ocmputational models, see data\simulated

File/folder	Description
simualted_data_PVLD_123	simulated data using the PVL-D model using seed 123
simualted_data_VSE_123	simulated data using the VSE model using seed 123

Scripts

All scripts are located under scripts subfolder

Data wrangling
scripts\1a_datawrangling contains code to

• Merge across all individual participant .txt data files
• Inspect and reformat data where needed
• Compute Iowa Gambling Task scores, by blocks
• Save merged data as Rdata file

Computational Modelling
Stan scripts for computational models can be located under scripts\stan-files.

File/folder	Description
1b_fit_model_loop	Loops to fit computational models
stan-files\PVL-D_sep_session.stan	PVL-D model stan script
stan-files\VSE_sep_session.stan	VSE model stan script
model_libary.R	Definition of corresponding stan files and parameters for PVL-D & VSE models

---

Note on data simulations

For our data simulation scripts, we included a single offset in the eta parameters in each model (i.e., Learning Rate parameters, or "Exploration Update" for VSE). These offsets create a "pre-" and "post-" value, assuming some sort of experimental manipulation.

See lines 42-43 for PVL-D simulation script and lines 45-46 for VSE simulation scripts. To simulate data with post-manipulation "changes" in other model parameters, you may replicate these lines accordingly with your desired effect size. In our scripts, we simulated a medium effect size (d = .40) using a mean of -.2 and SD of .5.