Figure 4 input data for "Local goodness-of-fit measures for neural decoding" (rat j16, 2021-07-10, run epoch 02_r1)

Derived input for Figure 4 of *Local goodness-of-fit measures for neural decoding*(Zeng, Comrie, Frank, Eden and Denovellis): the position and the spike-sortedhippocampal units of one run epoch (rat j16, 2021-07-10, epoch 02_r1) of thespatial-bandit recording of Comrie et al. (2026), *Neuron*,doi:10.1016/j.neuron.2026.08.023. The raw recording is on the DANDI Archive (dandiset 001942). The spike sorting usedhere is not currently on DANDI, so this file is what reproduces Figure 4 exactly. **Contents.** One NumPy `.npz` file of numeric and string arrays (no pickled objects;load it with `numpy.load(path, allow_pickle=False)`): - the position time series, 709,321 samples: head position, speed, velocity and orientation, linearized position and track segment;- the spike times of 203 units, 870,018 spikes (`spike_times`, split into units by `spike_offsets`);- the track graph used to linearize position. `statespacecheck_paper.load_local_data.recording_arrays` in the analysis code definesthe layout. **Integrity.** SHA-256 `60383b394b597e2900545548ecac7c53a8601038ace9dbeb42d7f9a5fe1c93b3`.The same value is recorded in the Figure 4 summary of the analysis code(`provenance.figure04_decode_cache.export_file_sha256`). **Use.** Put the file in `data/` folder of https://github.com/edeno/statespacecheck-paper andrun `uv run python scripts/generate_figure04.py`. **Provenance.** `docs/data-lineage.md` in the repository records the lab databaseentries the file was exported from and how that was verified.