-
-
Notifications
You must be signed in to change notification settings - Fork 395
Season of Docs 2021
We are thrilled to announce that ArviZ is one of the project accepted at Google Season of Docs 2021. We will publish the requirements and selection process this weekend (before April 19th). The project idea is already decided and is outlined below.
For further inquiries and questions on the project or budget described in this document, you can contact us on Gitter
ArviZ is a MIT Licensed library designed for the Exploratory Analysis of Bayesian Models in Python. Statisticians and data analysts, across both academic and industry, use ArviZ to assist them to perform statistical workflows to study a wide range of problems, from the effectiveness of cancer treatments, to SpaceX rocket supply chains, to user behavior on the internet. The folks using ArviZ as technical writers themselves, of some sort, as they are using ArviZ to translate mathematical results into “inference”, or in other words data driven conclusions. To aid with Bayesian modeling analysis ArviZ provides both numerical and visual summaries, as well as a collection of software tools that help with modeling and storage of statistical results.
ArviZ has two main challenges. One is that as a statistical tool it requires some familiarity with both programming and statistics which can present a challenge for newcomers. This barrier in entry makes it challenging for folks that are becoming interested to become confident and proficient. The other is that ArviZ is a collection of various methods and tools, and in practice there is no single workflow in which the tools are always used, but rather are chosen as needed. However this “toolbox” of methods makes it challenging for present all the tools available, let alone help folks understand which one is the right one in their particular situation.
A smoother on ramp into the knowledge captured in ArviZ’s docs, and in the contributors minds, would help users “level up” smoothly, understand all the mathematical and visualization tools available to them.
The Arviz documentation project will
- Evaluate current documentation flow and organization to understand the path to relevant information for various types of users
- Power users that know what they want
- Intermediate users that need to browse around to find what they need
- Beginners that are just starting out in statistics and are trying to get oriented
- Improve documentation readability ensuring the language is accessible by folks from any language background
- We believe this come in two levels. A guided form that is beginner friendly and reference manual for for advanced users
- Provide feedback on visual design of graphics
- Make code changes based on recommendations
- Include assessment of sister open source docs from the perspective of an ArviZ user such as xarray and matplotlib as these OSS libraries are crucial for use
Work that is out of scope for this project
- Extensive technical explanation of mathematical concepts. E.g. we’re not proving theorems or rewriting papers
- Long form tutorials. We’re not writing a book or case studies
The number of PyPI downloads increase by 10% from baseline set at start of project The number of visitors to the docs site increases by 15% from baseline set at start of project
- Pay graphic designer for new logo design $500
- Current design is a prototype from original package release
- Fund Code contributions related to project $500
- Some documentation may require code package changes
- Sister Project Funding $1000 -
- Aid downstream packages in building better documentation
- Technical Writer $7000
- Bulk of expense in evaluating documentation and providing us guidance on how to improvement, as well as direct improvements
- Visual designer for evaluation of visualization effectiveness (if needed) $2000
- Trained help in helping us understand if our packages plots are indeed effective in communicating the information they’re designed to
Total: $11,000