COURSES & OPEN LEARNING
Courses I currently teach, with links to their course sites.
Each site is written for enrolled students in Japanese, and collects the slides, notebooks, and assignments for every session.
Intuition and experience fail systematically in front of data. Students learn statistics as a shared language for describing and reasoning about phenomena without cognitive bias, checking every idea by running it in Python rather than calculating by hand.
Builds the foundations of statistical inference with conditional probability, Bayes’ theorem, and probability distributions. Students then use simulation, information cascades, and Condorcet’s jury theorem to show how conformity breaks the independence of trials and degrades group decisions.
Methods for checking social phenomena against data, covering both quantitative and qualitative research. The course moves from the history and ethics of research and the use of official statistics to questionnaires, interviews, observation, and ethnography.
A project-based course in which each student carries out one survey over the semester: choosing a theme, writing questions, collecting and analysing responses, and publishing the results as a web page and a report. It also covers how to ask generative AI and how to check its answers.
The basics of data analysis in the behavioural sciences, taught in R (RStudio): importing, tidying, and visualising data, then running and interpreting t-tests, ANOVA, and chi-square tests. No prior programming experience is assumed.
From digitising sound through spectrograms, acoustic features, statistical models, current recognition models and evaluation metrics. Every session carries a section on evolution, connecting that week's technology to how human speech and hearing came to be.
The first half covers the infrastructure: columnar storage, the CAP theorem, approximate algorithms. The second half turns to multivariate analysis through the objective each method optimises, working through multiple regression, principal component analysis, k-means and partial least squares.
Starting from the principles of visualisation, then data handling with DuckDB and Polars, visualising principal components and clusters, interactive charts, dashboard design and data storytelling. Each session ships an exercise script that offers graded hints while students solve it themselves.
Built on a reproducible analysis pipeline, covering coordinate transformation, spatial joins, nearest-neighbour analysis and kernel density estimation. The later sessions move to principal component analysis, k-means and partial least squares, and to reproducibility when generative AI enters the analysis.
The goal is not to memorise Word or Excel, but to decide what you want to do and to read and handle information correctly. The site keeps the materials for every session, along with NotebookLM slides, videos, and quizzes for review.
Eight ideas explained with primate data you can move, written for readers who never took the courses.
OpenWhere the same topic appears across five courses, and which course reaches it first.
OpenWhere each session's method sits among seven perspectives, with a child page reading them as formulas.
OpenAll course sites are in Japanese. A full teaching history is available on researchmap.