This repository contains all data and code associated with our manuscript: *Storage in visual working memory recruits a content-independent pointer system* Usage: If you would like to use the data in a published work, please cite both the paper (not yet published) and the OSF data set (see the top-right corner of the main project page for the citation). ## Getting started Some of these components contain a lot of data, so I recommend only downloading what you need. Each major analysis is contained in a iPython Notebook (`ipynb`). The `decode_eeg.py` script contains most of the major functionality. The notebooks import this script. The `mord` folder contains the package that allows for ordinal logistic regression classification. Alternatively, you can import and use the `sklearn.linear_model.LogisticRegression` which results in slightly worse decoding accuracy. To run the analyses, copy the `data/data` folder into the `analysis` directory. For analysis scripts, pulling from the GitHub might be easier. find the repo here: https://github.com/WilliamThyer/Thyer-et-al-2021 ## Analysis file descriptions `decode_eeg.py` - Contains vast majority of functionality for loading and wrangling data, cross-validation, classification, plotting, and stats testing. All notebooks import this package. - `Experiment` - Handles loading data. Directories, data info, subjects, etc. - `Experiment_Syncer` - Handles synchronizing subjects across multiple experiments using unique IDs. Also uses `Experiment` functions. - `Wrangler` - Handles data augmentation before classification. That includes trial binning, selecting classes, cross-validation, and rolling over time. - `Classification` - Handles actual classification. Basically standardizing data, training, and testing models. - `Interpreter` - Takes in results from classification. Handles all of the plotting and statistical testing of results. - `ERP` - Less connected than the other classes. Handles plotting of ERPs. `decode_eeg.ipynb` - Basic notebook for load classification within an experiment. Not actually used, since `decode_eeg_loop.ipynb` handles all within-experiment decoding. Still a useful template. `decode_eeg_loop.ipynb` - Loops through experiments and trial bin size parameter and performs classifications. `decode_load_single_feature.ipynb` - Trains classifiers on experiment 1 (color) and tests on experiment 2 (orientation) and vice versa. `decode_load_single_feature_set_size.ipyng` - Train classifiers on mixture of experiment 1 (color) and tests on experiment 2 (orientation) with set size 1 vs 2, 2 vs 3, & 3 vs 4. `decode_load_single_feature_to_conjunction.ipynb` - Trains classifiers on experiment 1 and 2 (single feature, color or orientation) and tests on experiment 3 (conjunction, color & orientation). `decode_preds_dont_double.ipynb` - Trains classifiers on experiment 1 and 2 (single feature, color or orientation) 1 vs 2 and 2 vs 4. Also trains classifiers on experiment 3 (conjunction, color & orientation) 1 vs 2 and 2 vs 4. Then compares predictions from those classifiers. `decode_load_acc_and_k.ipynb` - Classifies load on all unique subjects across all three experiments. Also calculates K for each subject. Then correlates classification accuracy and K. `decode_load_colinearity_hyperplane.ipynb` - Using data from Diaz et al. 2021, train classifier on set size 2 ungrouped and set size 4 ungrouped. Then test on set size 4 grouped (4 items, but arranged to look like 2). Measure distance from hyperplane for each condition. `behavior_analysis.ipynb` - Plots behavior. Specifically accuracy across set sizes. `plot_erp.ipynb` - Plots frontal, central, and parietal/occipital ERPs for each experiment. `decode_load_eyetracking.ipynb` - Classifies load in Experiment 1 using EOG data. Plots accuracy, setsize pair accuracy, and confusion matrices. ## Data file descriptions ### Data Contains the actual data files used for analysis. 1801, 1901, and 1902 are experiments 1 (color), 2 (orientation), and 3 (conjunction) respectively. `[exp]_[subject_number]_xdata.mat` - Contains the EEG data. Size (trials, electrodes, timepoints). `[exp]_[subject_number]_ydata.mat` - Contains trial set size labels. Size (trials). `[exp]_[subject_number]_behavior.csv` - Contains behavior data. `[exp]_[subject_number]_info.mat` - Contains EEG metadata. Unique ID, electrode labels, channel locations, sampling rate, and trial times. `[exp]_[subject_number]_eyetracking.mat` - Contains eye tracking data and labels. ### EEGLAB data Contains the EEGLAB formatted data. It is preprocessed and artifact rejected. If you want to do analyses in MATLAB, these are the files you should use. For further questions, you can email me at thyer@uchicago.edu