Reference · the portfolio

A panel reads for ten minutes. Give them evidence.

Almost every PhD applicant writes "proficient in R" on their CV, and it means nothing because it cannot be checked. Eight artefacts make the claim verifiable in the time a panel actually has: a manuscript that rebuilds itself, a repository they can clone, a preregistration with a power curve, a package they can install. Each one is an afternoon of assembly on top of work you have already done.

8
artefacts, each with a build day
1
page that links all of them
10
minutes of a reader's attention
{{ a.num }} {{ a.meta }}

{{ a.title }}

{{ a.body }}

What "done" means
  • {{ line }}
    •
What it signals
{{ a.signal }}
Interview question it invites: {{ a.question }}
{{ a.code }}
{{ a.taskKicker }}
{{ a.task }}
Assembly

The order to build them in

Each row feeds the next, so building them in this order means almost no duplicated work. Two focused weekends is a realistic total if the underlying analyses exist.

#BuildBecause it produces the input for
1Figure portfolio (day 24)Figures the manuscript and the workshop both need
2Analysis template with targets and renv (day 93)The infrastructure every other artefact sits on
3Your own package (day 95)Functions the template calls
4Reproducible manuscript (day 88)The centrepiece; uses 1, 2 and 3
5Public repository (day 91)Houses 1–4 and is the link on your CV
6Simulation study (day 84)Supplies the power curve for the preregistration
7Preregistration (day 96)Uses 6; demonstrates design thinking
8Workshop handout (day 96)Reuses everything above as teaching material
Assembly

The one page that links all eight

Scattered artefacts do not read as a portfolio. Put a single page — a GitHub profile README, a Quarto site, or one PDF — at the top of the pile, with one line and one link each. This is the thing you actually send.

--- title: "Methods portfolio — Your Name" format: html --- Doctoral applicant in clinical psychology. Quantitative methods work, all of it reproducible. Contact: you@university.edu · ORCID 0000-... ## Reproducible manuscript Experience-sampling study, 94 participants, 2,814 observations. Written in Quarto with papaja; every number in the text is computed at render time. [PDF](...) · [source](...) ## Analysis repository Cleaning, scoring, models, figures and manuscript in one `targets` pipeline. `renv::restore()` then `tar_make()` reproduces every result in four minutes. [github.com/you/stress-mood](...) ## Preregistration with simulation-based power Hypotheses, analysis plan, and a power curve simulated for the exact planned mixed model (80% power at 90 participants and 30 observations each). [OSF](...) · [power script](...) ## R package: labtools Scale scoring with explicit missing-data rules, a house ggplot2 theme, and importers for Qualtrics and m-Path exports. Documented, tested, R CMD check clean. `remotes::install_github("you/labtools")` [github.com/you/labtools](...) ## Figure portfolio Six figures at journal specification, each with the claim it supports. [PDF](...) · [script](...) ## Simulation study Coverage of the 95% interval when clustering is ignored, across nine conditions and 500 replications. [PDF](...) · [code](...) ## Workshop: multilevel models for diary data Two-hour Quarto tutorial with exercises and solutions, taught to the department's methods group in March 2026. [handout](...) · [slides](...)
Three rules for this page
Every link works and every repository runs. One broken link costs more credibility than a missing artefact. Clone each repo into a fresh folder and follow your own instructions before you send anything. One sentence of substance per item — the design, the n, the method — not adjectives. Nothing aspirational. An artefact that is half-built does not belong on the page; six finished items beat eight with two placeholders.
Assembly

The methods paragraph in your application

Most personal statements say "I have strong quantitative skills". Replace it with specifics and a link. Two versions below — the second is the same person, after this course.

Before
"I am proficient in R and SPSS and have experience with advanced statistical techniques including multilevel modelling and structural equation modelling. I am committed to open science practices."
Unverifiable, and identical to two hundred other applications.
After
"My analyses are fully reproducible: my master's study is a public repository where renv::restore() and tar_make() rebuild every figure, table and the manuscript itself from the raw data. I preregistered the design with a simulation-based power analysis for the exact mixed model I fitted, and I have written a small R package for the scale-scoring routines my lab reuses. All of it is linked at [url]."
Every claim is checkable in under ten minutes, which is why it is believed.

And if you are not applying anywhere

Build them anyway, for a reason that has nothing to do with applications: each artefact makes your next study faster. The template saves a month, the package saves a week per project, the figure set saves an afternoon each time, and the simulation habit means you never again collect data you cannot analyse. The portfolio is a by-product of working properly.