A 14-week graduate-style sequence for researchers who use statistics in their work and want to use them honestly. Taught by Solomon Ardent at Foundry 47 School since 2019.
A 4-hour 02-minute unit on what regression is and is not for. Eight lessons, replicated in R and Python, problem set 03 attached.
Each unit closes with a graded problem set; the fourth closes with a 6,000-word applied paper from your own work.
Reading, by hand, what a likelihood function actually says before any numerical optimisation. Problem set 01: derive three by hand, including a non-trivial mixture.
Why linear regression is the right answer 70% of the time and the wrong one when it is not. Coded in R; replicated in Python with statsmodels.
When your data are nested — students within schools, measurements within patients — and what stops working if you ignore it. Real datasets, not the radon one.
An applied paper using the methods of the course on a dataset from your own work. Reviewed twice — by Solomon and by a second-year cohort peer.
Multilevel survival models on a clinical dataset of 4,212 patients. The paper appeared in Northwind Bio Methods, March 2025. Soraya cited Unit III as the week the modelling stopped being intimidating.
An applied econometrics paper on regional policy interventions. Reviewed twice in cohort, then accepted at Mercury Econ Letters within four months of submission.
Hierarchical Bayesian models on hospital readmission rates across 47 sites. The paper became Maja's PhD chapter and was published in Public Health Methods, March 2026.
I am a working biologist, not a statistician. The Lectern course is the first time I felt I was being taught what I would actually use, in the order I needed it. I have stopped sending p-values to journals I do not believe.
Or write to solomon@foundry-47.edu — replies in 48 hours.
Eighteen researchers, fourteen weeks, one finished paper. Applications open in late autumn.