Personal Highlights: Positconf 2024

Data science

My picked talks for 2024 after Posit released the talks on YouTube

Author

Chi Zhang

Published

November 5, 2024

Quarto

Tyler Morgan-wall: Quarto, AI, and the Art of Getting Your Left Back

Transform old blog (Wordpress) to Quarto. Quarto can render pre-exisiting HTML

Customize website with AI, with css

Image carousel, transform some grid images into an auto rotating carousel (JS). Can also add screenshot; explain what the code is doing

Dashboards

Alenka Frim and Nic Crane: Mixing R, Python, and Quarto: Crafting the Perfect Open Source Cocktail

Comparing the tools to make dashboards:

Dashboard aesthetic Markdown syntax Deploy with GH actions Both R and python
Shiny Yes No No / Yes with Shinylive Yes
Quarto doc No Yes Yes Yes
Quarto dashboard Yes Yes Yes Yes

Engine: knitr and jupyter

Sean Nguyen: Beyond Dashboards: Dynamic Data Storytelling with Python, R, and Quarto Emails

Remove friction of too much dashboards

  • logging in can create barriers
  • meeting executives where they are (email, for example)
  • “no-click” insights (e.g. add key metric/alerts in the subject line or notification)

Tools: quarto emails, pins (a package), posit connect

  • data sources -> data warehouse (big query) -> quarto (create pins.qmd) -> pins (marketing data.csv pin, for example)
  • change quarto format to email
  • generate multiple using purrr

Python

Emily Riederer: Python Rgonomics

Python alternatives to R. Worth rewatching!

Teaching and education

Andrew Gard: Teaching and learning data science in the era of AI

Students don’t know enough to be able to edit the prompt to reach a sensible code chunk, AI guessed and guessed wrong. We should not expect AI to guess information that we do not provide!

Students should still learn to code, and teachers should ask better questions - instead of asking for the final result (create a bar plot), ask students to critically think: why doesn’t the AI-generated code work? what information is missing? how do you improve the prompt?

James Wade: Posit Academy in the Age of Generative AI - Lessons from the Frontlines

chattr, gptstudio, github copilot

Posit Academy learners (over half) give AI code assistants 2 star rating or less

Rewarding, high-growth period. Threshold concepts: once understood, transforms your perception and approach of a discipline, and these must be encountered not told.

TC in DS:

  • tidy data enables efficient analysis
  • modular code enhances re-usablity and clarity
  • visualization as a tool for exploration and communication

How to incorporate AI code assistants (in DS class)

  • early stage: explain this code piece by piece
  • mid stage: add a roxygen skeleton to my code
  • late stage: try code assistants in the IDE

TC for code assistants:

  • drive faster but don’t forget to steer
  • prompting matters, learning how to use these tools is a skill

Statistics

Hannah Frick: tidymodels for time-to-event data

Max Kuhn: Evaluating time-to-event models is hard

Demetri Pananos - Making sense of marginal effects