Ready4R (2024-09-02): Stats, Stats, (and GitHub)
A Continuation of my PositConf recap.
Welcome to the Weekly Ready for R mailing list! If you need Ready for R course info, it's here. Past newsletters are available here.
Hi All,
Continuing my recap of PositConf 2024 by talking about statistical applications and a great talk about GitHub and GitHub merges.
Keynote: Allen Downey, The Future of Data Science
In this talk, Allen Downey talked about the Past, Present, and Future of Data Science, noting that we are currently in the “trough of disillusionment” in terms of perceptions of Data Science. Data Scientist has been battered by bad applications, such as unfair algorithms. Still, Allen found reason to be more optimistic for the future, talking about society reaching realistic expectations for data science in general.
Allen Downey has a blog post called “There is Only One
Test” that was tremendously influential to me that talks about how with bootstrapping (drawing samples with replacement), you only need one statistical test. This is especially true given the computational power
available to us.
His slides are available here: https://tinyurl.com/dow24pos
{marginaleffects}
Demetri Pananos gave a demonstration of the {marginaleffects} package. One of the hardest parts of communicating models is describing the effects particular variables have on the prediction. {marginaleffects} lets you directly report average differences between groups on changing a covariate on the model in meaningful units.
For example, one meaningful difference might be the Risk Difference - what is the change in the predicted probability over a group when you change a variable? Demetri talked about the main steps in using {marginaleffects}, which are:
- Write Comparison as function (make a contrast)
- Specify Weights
- Use {marginaleffects} to make comparisons
The rest of this session (Machine Learning and Statistical Modeling) contained a number of really interesting talks, including “Understanding, Generating, and Evaluating Prediction Intervals” (Brian Shalloway) and “Keras 3: Deep Learning made easy” (Tomasz Kalinowski).
Please Let Me Merge Before I Start Crying
Meghan Harris’ talk was part of the session “Is it Supposed to Hurt this Much?” which focused on useful mechanics, such as Git, and working with Databases.
The full title of Meghan’s talk is “‘Please Let Me Merge Before I start Crying’ and Other Things I’ve Said at the Git Terminal”. This was a great introduction to more intermediate concepts of Git, including one of the most useful visual explanations of git merge and git merge.
Meghan did a great job demystifying how she works with git and making it more approachable for Data Scientists.
Check out her slides here:
https://meghansaha.github.io/please_let_me_merge/.
Signing Off
Well, that’s another recap newsletter. I think the last one for PositConf will be about the Teaching and Community Sessions. Thanks for reading this far!