Data and Analysis: Visualizing distributions of vehicle trajectories through naturalistic environments

Data and Analysis for Trajectory Visualization This includes data files and analysis notebooks for Visualizing distributions of vehicle trajectories through naturalistic environments by Benjamin T. Files, Laura R. Marusich, and Mark. S. Dennison Ben Files is the point of contact: [e-mail] Project Root The project root has `_export.ipynb` files to convert the csv data into json for cmdstan to use. The bash scripts run_exp1.sh and run_exp2.sh were used to submit jobs to our high-performance computer center, so there are a bunch of slurm-specific things in there, but the main point of the script is to run the model with `cmdstan`. Results summaries, figures, and statistics are generated in `exp1_results-zoib_pinf_presentation.ipynb` and `exp2-2_results-zoib_pinf_presentation.ipynb`. Those make use of `zoib_util.py` for fit diagnostics, but the other functions in there are not used. Data The `data` folder includes `TrajViz1_data_complete.csv` and `TrajViz2_data_complete.csv`, which are exports from the database where the original data are stored. This is the most convenient starting point. Also in the `data` folder are .json files that package the data from their respective .csv files into a format that is usable by `cmdstan`. These files are generated in `exp1_export.ipynb` and `exp2_export.ipynb` in the project root. Stan Models The `stan_models` folder contains the stan model, `zoibeta_pinf_full_shard.stan`. This model uses within-chain parallelization to speed up sampling, and some of the data transformations were done to facilitate this. Sampling was done on a node at a high performance computing center. More typical hardware might not benefit from (and might even be slowed down by) this parallelization. Stan Output Under the `stan_output` folder are samples generated by `cmdstan` with our model and data. Note: We included seeds in our sampling scripts, but that is not enough to guarantee full reproducibility. See https://mc-stan.org/docs/reference-manual/reproducibility.html Prerequisites Python packages: pandas, numpy, arviz (pre-1.0 version), xarray, plotnine, matplotlib. jupyterlab for the notebooks. To run the sampling, a configured and working installation of cmdstan is required. License This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.