# Code for: <MANUSCRIPT TITLE>

<!--
  README template from the TADA-Wiki.
  Covers every field required by the "Documented" TADA guideline and by
  Stages 1.5 and 3.5 of the SORTEE data and code quality control guidelines.

  Replace everything in <ANGLE BRACKETS>, delete what does not apply, and
  delete these comments before archiving.

  Under double-blind review you may leave the Contact and Funding sections
  blank, and fill them in on acceptance.
-->

Associated manuscript: <AUTHORS> (<YEAR>), *<JOURNAL>*.
<PREPRINT OR ARTICLE DOI>

## Abstract

<Paste the manuscript abstract here.>

## Contact

<NAME> — <EMAIL>
<!-- Add ORCID if you have one. -->

## Funding

<FUNDER>, grant <NUMBER>.

## Licence

Code: <e.g. MIT — see LICENSE>
Data: <e.g. CC0 / CC BY 4.0>

## Data

<Where the data live. If archived with this project, say so. If archived
separately, give the persistent identifier:>

  DOI: <https://doi.org/...>
  Files: <name and one-line description of each data file>

<If any data are derived from other sources — other articles, online
databases — say which, and cite them.>

Collected <WHEN> at <WHERE>.

<!--
  Every variable in every data file must be described, with units. Do this
  here, in a separate data README, or in a data dictionary (a .csv with a
  column for variable names and a column for descriptions).
-->

### Variables in `<datafile.csv>`

| Column | Description | Units / levels |
|---|---|---|
| `<name>` | <what it is> | <unit, or the set of possible values> |

## Software environment

<LANGUAGE> v<VERSION>   <!-- e.g. R v4.3.3 -->

Key packages:

  - <package>   v<version>   <what it is used for>
  - <package>   v<version>   <what it is used for>

<Full environment in `session_info.txt`, produced with `sessionInfo()` in R or
`session_info` in Python.>

<!-- State whether and how large language models were used in code generation. -->
LLM use: <none / describe>

## How to run

<Say how the project should be opened: open `<project>.Rproj` in RStudio;
activate the virtual environment; open the folder in VS Code. This matters —
it is what makes the relative file paths resolve.>

Run the scripts in this order:

  1. `<01_script.R>`   reads  <input>
                       writes <output>
                       <what it does>

  2. `<02_script.R>`   reads  <input>
                       writes <output>
                       <what it does>   (~<runtime>)

  3. `<03_script.R>`   produces <Figure X, Table Y>

Total runtime approximately <TIME> on <HARDWARE>.

<Flag anything slow, memory-hungry, or requiring specialist hardware. If a step
is expensive, say whether its output is cached and where.>

<If any step needs proprietary software, say which parts can and cannot be run
without it, and archive the saved outputs.>

## Files

  data/raw/         <Unmodified. Never written to by any script.>
  data/processed/   <Generated by script 01. Safe to delete and regenerate.>
  R/                <Analysis scripts, numbered in run order.>
  functions/        <Helper functions sourced by the scripts.>
  output/           <Figures and saved model objects.>

## Results in the manuscript

<Optional but very useful: a map from manuscript output to code.>

| Manuscript | Produced by |
|---|---|
| Table 1 | `<script>`, section `<name>` |
| Figure 2 | `<script>`, section `<name>` |

## Reproducibility notes

<Anything a reader needs in order to interpret a mismatch:>

  - Seeds are set at the top of `<script>` before <stochastic step>.
  - <Figures were post-processed in <software>; the underlying data are
    reproduced exactly but the rendering may differ.>
  - <Results are sensitive to <package> version <X> because <reason>.>