Learner Profiles

Who Should Take This Workshop?

Workshop 8 — R for Reproducible Scientific Analysis — assumes you know basic R syntax (variables, vectors, functions, control flow) and now want to organise that knowledge into project-shaped, reproducible analyses you can hand to a journal, a co-author, or your future self. The workshop leans heavily on the tidyverse (dplyr, tidyr, ggplot2), R Markdown, and the here/renv ecosystem.

The three personas below are typical of who enrolls. They all use the Sagehen HPC cluster either as their primary R environment (via the OnDemand RStudio Server app at https://ondemand.hpc.pomona.edu) or as the place where their long-running analyses end up. Wherever the lesson refers to “your R session,” you can read “the OnDemand RStudio Server session running on Sagehen’s amd partition.”

If you have not used R before at all, take Workshop 7 (Programming with R) first.

Profile 1: Dr. Lin — Postdoc Preparing a Reproducible Analysis for Journal Submission

Background

Dr. Lin (she/her) is a second-year postdoctoral fellow in the Pomona College psychology department, working on a longitudinal study of executive function in undergraduate athletes. She is preparing to submit her first solo-authored manuscript to Psychological Science, which now requires a deposit of fully reproducible analysis code. Her PI’s previous lab worked in SPSS; she taught herself R during her PhD and has been writing increasingly long single-file scripts since then. She knows the analyses are correct because she has triple-checked them, but she also knows her code is “spaghetti”: one 800-line script, manual file paths, no project structure, and figures pasted into Word.

What She Knows

  • Tidyverse basics (dplyr::filter, dplyr::mutate, ggplot2)
  • Linear mixed-effects models with lme4
  • read_csv, write_csv and basic plotting
  • How to write a function (with one or two arguments)
  • The general shape of an R Markdown document (she has knit one once)

What She Doesn’t Know Yet

  • How here::here() makes a script portable across machines (and across her laptop and Sagehen HPC)
  • How to split a long script into logical R Markdown sections that knit reproducibly end to end
  • How renv records package versions so a reviewer running her code two years from now gets the same numbers
  • How to write a Makefile (or targets pipeline) so the figures regenerate automatically when the data changes
  • How to publish her work as a GitHub repo + Zenodo DOI

Why She Needs This

  • Her journal requires it. Psychological Science’s reproducibility checklist is real and is checked.
  • Her PI is watching. If she nails this submission, she goes on the next R01 as a co-investigator.
  • Her data is on Sagehen. The longitudinal data lives at /bigdata/lab/psychlab/longitudinal2024/ and a full reanalysis takes ~6 hours — too long for a laptop, but a sensible SLURM batch job on the amd partition.

How She Will Use Sagehen HPC

Dr. Lin runs interactive analyses on the Sagehen OnDemand RStudio Server (8 cores, 64 GB RAM), but submits the final knit as a SLURM batch job (sbatch knit_paper.sh, which calls Rscript -e "rmarkdown::render('paper.Rmd')"). All data lives under /bigdata/lab/psychlab/; her R code lives in a Git repo synced to Pomona’s institutional GitHub.

Success Indicator

By the end of the workshop, Dr. Lin can:

  • Restructure her 800-line script into a project with data/, R/, reports/, and figures/ subdirectories
  • Use here::here() so her code runs unchanged on her MacBook and on Sagehen
  • Knit a single R Markdown file end to end that produces every figure and number in the manuscript
  • Capture her exact package versions in a renv.lock file
  • Push the repository to GitHub and tag a v1.0-submission release

How to Pace for Dr. Lin

She is sophisticated but anxious; she wants to know that the things you are teaching are the minimum she needs to satisfy Psychological Science, and she will push back on anything that feels like polish-for-its-own-sake. Tie every concept to her submission deadline.

Profile 2: Professor Alvarez — Faculty Member Teaching Reproducible Methods

Background

Professor Alvarez (he/him) is a tenured associate professor in Pomona’s economics department. He has used R for fifteen years but mostly as a “fancy calculator”: one script per chapter, setwd() calls scattered everywhere, copies of the data in three folders. Pomona has just adopted a new methods requirement for the economics major: every senior thesis must include a fully reproducible analysis. Professor Alvarez has been volunteered (kindly but firmly) to teach the new junior-year methods course in the spring. He wants to learn the modern reproducible-analysis stack himself before he asks 22 anxious juniors to learn it from him.

What He Knows

  • A great deal of base R and statistical methods
  • Stata, SAS, and Eviews from earlier in his career
  • LaTeX and Beamer
  • The basic shape of GitHub (he has a personal account; he has never opened a pull request)
  • Excellent statistical and pedagogical instincts

What He Doesn’t Know Yet

  • The tidyverse — he has read about dplyr but always reaches for apply and [
  • R Markdown — he knit one document in 2017 and gave up
  • How to teach students to use here::here() instead of setwd("C:/Users/Carlos/Desktop/")
  • How to use Sagehen for class assignments (each student needs a working environment)
  • renv — he is willing to learn it but skeptical of “yet another package manager”

Why He Needs This

  • He is teaching this in the spring. The course is on the books; the syllabus is due in eight weeks.
  • His students will use Sagehen. With 22 students all needing R 4.5.3, RStudio, and identical package versions, the OnDemand RStudio Server on Sagehen is the only sane delivery model.
  • He cares about his discipline. Several recent retraction scandals in economics have involved unreproducible Stata code; he genuinely wants his students to do better.

How He Will Use Sagehen HPC

Professor Alvarez will set up a course shared directory at /bigdata/lab/economics/methods201/ with the IPUMS-CPS extracts and the assignment templates. Each student will have their own ~/methods201-USERNAME/ Git-backed project directory. He’ll show them how to launch RStudio Server through OnDemand on the amd partition with modest resources (--cpus-per-task=2 --mem=8G --time=4:00:00).

Success Indicator

By the end of the workshop, Professor Alvarez can:

  • Convert one of his recent papers’ analyses into a tidyverse pipeline + R Markdown report
  • Use here::here() and explain why it matters (in language a junior with a stats minor will understand)
  • Capture a course-wide renv.lock so all 22 students start the semester on identical package versions
  • Walk a student through cloning the assignment repo on Sagehen, running the report, and pushing back changes
  • Articulate the “minimum reproducible unit” he will require for each assignment

How to Pace for Professor Alvarez

He has a pedagogical lens on everything — every concept needs to be teachable, not just learnable. Pause regularly to ask, “How would you explain this to a junior?” He will appreciate the question even if he doesn’t always have an answer.

Profile 3: Sam — Staff Statistician Supporting Multiple Research Groups

Background

Sam (they/them) is a research statistician on Pomona’s Research Computing team, embedded part-time across the chemistry, biology, and neuroscience departments. They have a master’s in biostatistics and have been programming professionally for ten years — initially in Stata and SAS, then increasingly in R. Their job is to drop into a lab with a panicked PI three days before a grant deadline and make the analysis work. The variety is the hard part: in any given month, Sam might write a meta-analysis pipeline for chemistry, a survival-analysis Shiny app for biology, and a Bayesian power calculation for neuroscience. Each group has its own data conventions and its own definition of “the analysis is done.”

What They Know

  • Excellent base R, tidyverse, data.table, and Stan
  • Project management and time estimation under pressure
  • Git, GitHub, GitLab, and pull requests
  • Linux, SLURM, and Sagehen’s storage layout (they completed Workshops 0–2 in 2023)
  • Several reproducibility frameworks — they know renv, groundhog, and targets exist but don’t yet have strong opinions

What They Don’t Know Yet

  • Which of those reproducibility tools they should standardise across the labs they support
  • How to set up a turnkey “new project” template that any of their PIs can spin up on Sagehen in five minutes
  • How to teach R Markdown to a wet-lab PI who has never opened a code editor
  • The right way to package up a finished analysis for archival on Pomona’s institutional repository or Zenodo

Why They Need This

  • They want a standard. Right now every project they hand off has slightly different conventions. They want a single “this is how we do reproducible analysis at Pomona” pattern they can offer to every lab.
  • They want a teaching artifact. The same template, polished, becomes the foundation for the workshops and office hours they run for graduate students and postdocs.
  • They want to retire technical debt. Their three-year archive of analyses-for-other-people is full of setwd calls and unversioned packages. Modernising the older ones in the next six months is a stated goal in their performance review.

How They Will Use Sagehen HPC

Sam already lives on Sagehen. Their typical day is an interactive session on the amd partition (16 cores, 64 GB), with longer-running Stan models or simulations submitted as SLURM batch jobs. They use /bigdata/lab/sam/ as their staging area; finished analyses migrate to the relevant lab’s /bigdata/lab/<labname>/ directory.

Success Indicator

By the end of the workshop, Sam can:

  • Articulate a single, defensible Pomona-standard project structure (folders, here::here(), renv, R Markdown)
  • Create a usethis::create_package() or cookiecutter-style template repository other labs can fork
  • Hand a wet-lab PI a one-page “How to run my analysis on Sagehen” cheat sheet that actually works
  • Submit a knit-the-paper SLURM job for a former client and verify it produces identical figures
  • Write a short memo to the Pomona ITS director recommending which tools to standardise on

How to Pace for Sam

They will be the most advanced learner in the room. Engage them as a co-teacher: ask them how they explain here::here() to a chemistry PI, and have them critique the workshop’s recommendations. They will bring the others up faster than you can.

Common Threads

Despite very different career stages, all three personas share four things:

  1. They already know R. This is not a “first programming language” workshop. If you are not comfortable with vectors, functions, and control flow, take Workshop 7 first.
  2. Their stakes are real. A journal submission, a syllabus deadline, a portfolio of half-finished analyses. Pacing matters; pure-theory tangents do not.
  3. They will use Sagehen. Long-running analyses, shared course directories, and large data on /bigdata push every one of them onto the cluster eventually. Treat OnDemand RStudio Server as the default environment, not an afterthought.
  4. They want a defensible standard. Each of them is, in their own way, trying to retire technical debt. The workshop’s success is measured in how confident they leave about what a reproducible analysis looks like, not just how to write one.

What Comes Next

After Workshop 8, all three personas are well prepared for:

  • Workshop 9 — SLURM Job Scheduling on Sagehen, particularly relevant for Dr. Lin and Sam, whose pipelines belong in batch jobs
  • Workshop 11 — OnDemand Portal Orientation, which Professor Alvarez should take and then assign to his methods students
  • The Pomona Reproducibility Office Hours run by Sam, which Dr. Lin will attend during the manuscript-revision phase

Cecil Sagehen approves of all of the above.