JapanStatsMAPs: a research map for Japanese public statistics
Contents
I have launched JapanStatsMAPs, a website for finding research papers that use Japanese official statistics. The data and code are available on GitHub.
As of November 7, 2026, the site indexes 1,269 papers, which you can search by theme, data source, analyzed data year, and study design. This article explains the site's contents, how to use it, the classification methods, and points to keep in mind.
About the site
Official statistics include published aggregate tables that anyone can download—such as the NDB Open Data and Vital Statistics—as well as microdata that require an application, like the Comprehensive Survey of Living Conditions. This site is a knowledge map collecting Japanese studies that used these statistics in their primary analysis.
No registration is required. By combining data sources and research themes, you can check which statistics have been analyzed for which years and identify under-researched combinations.
Introductory video
I created a narrated video of about 50 seconds showing how to navigate the site.
The narration uses text-to-speech. I used the "Ayaka" voice from the ElevenLabs Voice Library with the Eleven v4 speech synthesis model.
Included papers
The site includes papers whose primary analysis uses Japanese data falling into either of the following categories.
| Category | Description | Count |
|---|---|---|
| A: Public aggregate data | Published aggregate tables anyone can download (e.g., NDB Open Data, e-Stat statistical tables, published Vital Statistics tables, National Police Agency suicide statistics) | 812 |
| B: Government survey microdata | Individual-level records from government statistical surveys obtained through secondary-use applications (e.g., Comprehensive Survey of Living Conditions, National Health and Nutrition Survey, Patient Survey, Vital Statistics death records) | 445 |
| A+B: Combined | Studies combining A and B | 12 |
The paper counts are as of November 7, 2026. The site excludes specially extracted NDB data, DPC data, commercial claims data, hospital records, and independent cohorts or surveys. It also excludes papers that use published figures solely as population denominators or standards for age adjustment while relying on other datasets for their primary analysis.
The search covers 27 data sources. These include the Vital Statistics, Comprehensive Survey of Living Conditions, National Health and Nutrition Survey, NDB Open Data, Survey of Medical Institutions / Statistics of Physicians, Dentists and Pharmacists, National Epidemiological Surveillance of Infectious Diseases, Population Census, and Labour Force Survey.
What you can do across the six views
| View | Features |
|---|---|
| Concept map | Displays themes as circles and "exposure → outcome" pairs identified in the papers as arrows. Circle size and arrow thickness represent the number of papers. |
| Association matrix | Shows the number of papers for each exposure–outcome pair in a table to help find under-researched combinations. |
| Data sources and years | Uses color shading to show which data years have been analyzed for each statistical source. |
| Paper list | Search by author, title, journal, or DOI, and filter by category, data source, and data year. Results can be saved as a CSV file. |
| Paper network | Places a selected paper near other papers that share references or themes. |
| Statistics and data | Check the number of collected and screened papers, breakdowns for each item, and the methodology. |
There are 65 themes, displayed across 14 groups.
Example workflow
Here is an example from the site's "How to use" section for starting a study on regional disparities in suicide.
- Click the "Suicide and self-harm" circle on the concept map to see which data sources have been used.
- Check the "Suicide and self-harm" column in the association matrix to find themes rarely studied as exposures.
- Enter "suicide" in the paper list, filter the data source to "Vital Statistics," and read the original papers.
- Click "View connections" on a useful paper to trace related studies.
How papers are collected and classified
I collected the papers by searching Europe PMC across the full text using the query "English name of the data source AND (Japan OR Japanese)". Because the name of a data source often appears only in the methods section, searching titles and abstracts alone misses relevant papers.
For classification, I used TypeSafe's evaluation model, Jev. The pipeline first screens titles and abstracts to assign categories, and then evaluates candidate papers in detail. The detailed evaluation receives the full-text methods and results sections for open-access papers, and the title and abstract for all others. As of November 7, 2026, the pipeline had collected 8,460 papers, identified 1,970 candidates, and included 1,269 papers. It excluded 71 papers after determining that their primary analysis did not use Japanese data.
As shown in the update log, the system searches Europe PMC daily for papers registered through the previous day and adds matching studies. Similarity between papers is calculated from shared references retrieved via OpenAlex and overlapping themes.
Caveats for users
- Classifications are automated and have not been manually verified. Because they may contain errors, always check the original paper before citing or making decisions.
- Arrows indicate analyzed pairs and do not imply causality.
- Blank cells or small theme nodes do not necessarily mean that no research exists for that combination.
- Because collection relies on English search terms, papers in Japanese-language journals not indexed in Europe PMC are excluded.
- For papers evaluated using only the title and abstract, the system may miss data sources or analyzed data years mentioned exclusively in the methods section.
What is available on GitHub
The repository contains the website itself, the indexed dataset, classification results, item definitions, and scripts to rebuild the project. The code and data (classification results, schemas, and interface) are licensed under CC BY 4.0. Bibliographic metadata such as paper titles and author names belong to their respective publishers and authors. Full-text files and raw outputs from Jev are excluded from the repository due to file size and copyright restrictions.
This work was supported in part by JSPS KAKENHI Grant Number JP23K16359 (Grant-in-Aid for Early-Career Scientists: "Analysis of suicide factors during the COVID-19 pandemic using an all-municipality prediction model").