## Introduction
This data release corresponds to a 2022 reanalysis of the NOvA 2020 $\nu_e$ appearance and $\nu_{\mu}$ disappearance data (data first analyzed in [Phys. Rev. **D**106, 032004](https://doi.org/10.1103/PhysRevD.106.032004)).
The reanalysis uses a Bayesian statistical approach instead of the preceding frequentist one.
Full details on the reanalysis may be found in the publication: [arXiv:2311.07835](https://arxiv.org/abs/2311.07835).
## Assumptions
The exposure, selections, and reconstruction are the same as in the 2020 paper cited above.
Markov Chain Monte Carlo was used to generate samples from the model. See the paper for more details.
## Data release contents
### Posteriors and credible regions
There are four files containing ROOT `TH1`s and `TH2`s (corresponding to posterior probability densities in the labeled variables) and `TGraph`s (corresponding to credible interval contours containing various fixed total probabilities in the 2D posteriors).
These can be used to recreate the posteriors shown in the paper.
Key for decoding the objects in these files:
* **Filenames** are formatted as `posteriors_{wrc|worc}_{indep|joint}-ord-marg.root`
* `wrc` --> "with reactor constraint"; `worc` --> "without reactor constraint" (on $\sin^2 2\theta_{13})
* `joint` --> marginalized over both mass orderings simultaneously, as in the paper cited above; `indep` --> marginalized over each mass ordering separately (as in the supplemental material to the paper).
* **Objects in files** are named with a long prefix shared amongst all objects and a suffix like `_{NO|IO|BO}_[cred_int_{06287|09545|09973}]`:
* `` will be the variable name (or two variables concatenated together if 2D):
* `dm32` --> $\Delta m_{32}^2$
* `ssth23` --> $\sin^2 \theta_{23}$
* `dcp` --> $\delta_{CP}$
* `J` --> Jarlskog invariant
* `NO` --> normal ordering, `IO` --> inverted ordering, `BO` --> both orderings
* The number following `cred_int`, if it is present, indicates the fraction of the posterior contained by the credible region: 62.87%, 95.45%, or 99.73%.
### MCMC samples
The file `arxiv-2311.07835.data-release.mcmcsamples.root` contains approximately 68 million MCMC samples corresponding to the Stan MCMC sampler run with the model and data described in the paper.
The samples here are implicitly marginalized across all the systematic uncertainty degrees of freedom used in the model but do not explicitly sample them.
These samples can be used to reproduce the plots in the paper.
Please consult the included example Jupyter notebook `NOvABayes2022DataRElease_ExampleUsage.ipynb` for examples of how the plots in the paper could be reproduced, along with notes about common pitfalls to avoid.