Sign in

Erika

@erikaduan.bsky.social
86 followers 157 following 6 posts

Public sector data scientist and former immunologist. I'm just here for the #rstats

PostsRepliesMedia
Reposted by Erika
Emil Hvitfeldt @emilhvitfeldt.bsky.social · 20/08/2026
I'm done drafting another section of "Feature Engineering A-Z", this section covers all the up and down sampling methods that people try to use to deal with imbalanced data. I'm covering all methods that people use, not necessarily the methods you have to use feaz-book.com/imbalanced #dataBS
feaz-book.com
Feature Engineering A-Z | Imbalanced Overview
Whether to keep this chapter in this book or not was considered, as the methods in this section are borderline feature engineering methods.
1121
Reposted by Erika
Jumping Rivers @jumpingrivers.com · 20/07/2026
AI in Production 2026 Day 2 talks are now on YouTube. A full day of practitioners on running machine learning in production. Watch: www.youtube.com/playlist?lis... Registration for 2027 is also open (10–11 June, Newcastle): ai-in-production.jumpingrivers.com/
071
Reposted by Erika
Andrew Heiss @andrew.heiss.phd · 02/05/2026
Got to present this early draft paper today at a neat workshop organized by @rohanalexander.bsky.social. It's my attempt to grapple with the need for struggle and friction in learning in a world where LLMs remove all that friction. Read the whole thing at stats.andrewheiss.com/pinky-lychee/
AI for Me, Not (Yet) for Thee? Desirable Difficulties and Deliberate Friction with LLMs

Author: Andrew Heiss

Affiliation: Andrew Young School of Policy Studies, Georgia State University

Latest version: May 1, 2026

Details: This is a short position paper written for the Toronto Data Workshop’s 2nd Colloquium on Applications of AI, “Notes on the Future of Quantitative Social Science,” held at the University of Toronto on May 1, 2026.

Abstract: I am simultaneously a critical skeptic of large language models (LLMs) and their role in the learning process, and a reluctant and wary user of those same tools in my own work. This essay is an attempt to work through this tension by examining how LLM-based tools shape and reshape the foundations of expertise required to use them. If learners can generate working code through a Google search, if they can upload a blank problem set template and get a completed assignment in seconds, and if they can design, carry out, and analyze research projects in minutes, how can they build the expertise that is required to work with LLM output? Relatedly, if experienced researchers spend the bulk of their time reviewing LLM output and not working through code themselves, what happens to their expertise? The answer to these questions, I argue, is to deliberately introduce difficulty and friction in quantitative work, allowing learners to build expertise and helping experienced researchers maintain it.
69830
Reposted by Erika
Olivia Guest · Ολίβια Γκεστ @olivia.science · 02/04/2026
jonny over on masto doing god's work: > anthropic [is running] an ENTIRE AD CAMPAIGN around "Claude code is written with claude code" and after the source was leaked that has got to be the funniest self-own in the history of advertising because OH BOY IT SHOWS. neuromatch.social/@jonny/11632...
neuromatch.social
jonny (good kind) (@jonny@neuromatch.social)
Attached: 1 image reminder that anthropic ran (and is still running) an ENTIRE AD CAMPAIGN around "Claude code is written with claude code" and after the source was leaked that has got to be the funn...
21835284
Erika @erikaduan.bsky.social · 02/04/2026
Things I wish I could show past me. (Some speakers at an early career seminar I attended as a PhD student once told the entire audience that the only way to make it in science was to publish at least 1 first author Nature paper.)
020
Reposted by Erika
Richard McElreath 🐈‍⬛ @rmcelreath.bsky.social · 22/03/2026
Statistical Rethinking 2026 is done: 20 new lectures emphasizing logical and critical statistical workflow, from basics of probability theory to causal inference to reliable computation to sensitivity. It's all free, made just for you. Lecture list and links: github.com/rmcelreath/s...
12683216
Reposted by Erika
Vincent Arel-Bundock @vincentab.bsky.social · 12/02/2026
"Model to Meaning" just received a super generous review in JASA. Check it out! (I'm blushing 😊) Reminder: You can get the paper copy from CRC, and the full **free** version will remain online forever at marginaleffects.com www.tandfonline.com/doi/full/10....
tandfonline.com
Model to Meaning: How to Interpret Statistical Models with R and Python
Published in Journal of the American Statistical Association (Just accepted, 2026)
67411
Reposted by Erika
Iris Meredith @iris-meredith.bsky.social · 13/12/2025
Dark times call for desperate measures, and with my career being a little hosed, I'm writing a book! It's about teaching R users to write code like software engineers do, and the hope is that I can use it partially as content marketing for my coaching offerings. Here's the first chapter: #R #R-lang
deadsimpletech.com
R the Software Engineering Way: Introduction and Chapter Zero | deadSimpleTech
It is worth noting from the very beginning that a software engineer's work doesn't start with writing code, but with setting up the development environment and the tools that they need to write code e...
27922
Reposted by Erika
Saloni @scientificdiscovery.dev · 09/12/2025
Big new blogpost! My guide to data visualization, which includes a very long table of contents, tons of charts, and more. --> Why data visualization matters and how to make charts more effective, clear, transparent, and sometimes, beautiful. www.scientificdiscovery.dev/p/salonis-gu...
screenshot of my post
28802312
Reposted by Erika
Emi Tanaka @emitanaka.org · 23/11/2025
My paper "Examining the Interface Design of Tidyverse" is now published in ANZJS. This paper calls attention to the importance of UI/UX principles & approaches to statisiticians and #rstats developers. Many thanks to reviewers who made this paper better! 🔗 onlinelibrary.wiley.com/share/author...
the UI design principles and the two UI approaches
2428
Reposted by Erika
Bruno Rodrigues @brodriguesco.bsky.social · 29/10/2025
new blog post: Of course, someone has to write imperative code to build reproducible data science pipelines. It doesn’t have to be you. brodrigues.co/posts/2025-1...
0319
Reposted by Erika
Richard McElreath 🐈‍⬛ @rmcelreath.bsky.social · 15/10/2025
Looks like a good guide - the general data cleaning part is a lean intro to some very common issues in all sorts of data. Would be great if every phd who touches raw data was offered a short course in these basics (in R or Python or whatever HipsterScript) cleaning-data-r.ala.org.au/2_general-cl...
cleaning-data-r.ala.org.au
3  Column classes & names – Cleaning Biodiversity Data in R
24410
Reposted by Erika
Richard McElreath 🐈‍⬛ @rmcelreath.bsky.social · 01/10/2025
Was asked about collinearity again, so here's Vahove's 2019 post on why it isn't a problem that needs a solution. Design the model(s) to answer a formal question and free your mind janhove.github.io/posts/2019-0...

tl;dr

    Collinearity is a form of lack of information that is appropriately reflected in the output of your statistical model.
    When collinearity is associated with interpretational difficulties, these difficulties aren’t caused by the collinearity itself. Rather, they reveal that the model was poorly specified (in that it answers a question different to the one of interest), that the analyst overly focuses on significance rather than estimates and the uncertainty about them or that the analyst took a mental shortcut in interpreting the model that could’ve also led them astray in the absence of collinearity.
    If you do decide to “deal with” collinearity, make sure you can still answer the question of interest.
311336
Reposted by Erika
Julia Silge @juliasilge.com · 10/08/2025
I'm excited to speak this afternoon at #useR2025 on outgrowing your laptop with #Positron for #rstats users! You can check out my slides at juliasilge.github.io/useR-2025/
Galaxy brain meme format outlining options for working with data: CSV file, parquet & duckdb, databases, and remote SSH sessions
15514
Reposted by Erika
Noah Greifer @noahgreifer.bsky.social · 16/04/2025
Really nice new paper by Jingyu Zhang, Oliver Lüdtke, and Alexander Robitzsch on the performance of doubly robust estimators of the ATE. A great example of clear writing and reporting, useful visualization through tables, and a review of modern literature. osf.io/5uj2f_v2
osf.io
OSF
23511
Erika @erikaduan.bsky.social · 13/07/2025
I've written a simple guide to the new Positron IDE for #rstats and #python programming. I think that RStudio is still the most thoughtfully designed IDE for R programming but Positron is very useful if you also code a lot in Python. github.com/erikaduan/r_...
github.com
2254