Software

R packages published on GitHub, plus browser-based HTML apps and R Shiny apps that run in the browser with webR — no installation required.

R packages

nmfkc

Non-Negative Matrix Factorization with Kernel Covariates. Available on CRAN. Featured in R-bloggers' "April 2026 Top 40 New CRAN Packages".

Docs GitHub CRAN R-bloggers

ljmds

Local Jaccard MDS for Longitudinal Binary Data.

Docs GitHub

nolap

Non-overlapping arrangement of scatter-plot points.

Docs GitHub

pord

Exact conditional tests for paired ordinal data recorded on the same K-point scale (e.g. expectation vs. achievement). A generalization of Fisher's exact test conditioning on both margins, with the exact null distribution obtained by dynamic programming.

GitHub

HTML apps

All run without installation. Your data is processed in the browser and never uploaded. Each app has its own Japanese/English toggle.

The provenance and licences of the demo data bundled with each app are listed in THIRD-PARTY-DATA.md.

Text & DFM analysis

From text to visuals — connected by one CSV

First, the Text analysis app builds a document-feature matrix (DFM) from your text and exports it as CSV. That same CSV loads directly into the three apps below — taking you from raw text to word clouds, co-occurrence networks, biplots, and group-wise topics without writing any code.

Text analysis (build DFM) → ① DFM visualization ② Correspondence analysis ③ NMF · Topic model

Text analysis (Document-Feature Matrix)

Tokenize Japanese text and build a document-feature matrix (DFM). Detect collocations, remove or replace words, and export to CSV.

Open app GitHub

DFM visualization (word cloud & co-occurrence network)

Loads the DFM CSV exported by the Text analysis app and draws a word cloud and a co-occurrence network. Adjust the number of words shown, thresholds, and community-detection methods interactively.

Open app GitHub

Correspondence analysis

Loads a DFM with attribute columns (from the Text analysis app), applies correspondence analysis to the attribute × word crosstab, and draws a biplot placing categories and words on the same plane.

Open app GitHub

Non-negative Matrix Factorization (NMF · Topic model)

Runs non-negative matrix factorization (NMF). The Topic model tab reads a DFM with attribute columns (from the Text analysis app) and shows topic composition by group; the Matrix factorization tab takes any matrix and shows heatmaps & two-way clustering.

Open app GitHub

Regression & multivariate analysis

Principal component analysis (biplot)

Eigen-decomposes the correlation matrix (standardize on) or the covariance matrix to obtain the principal components. Reports the importance of components, a scree plot, loadings and scores, and draws a biplot placing the scores of the individuals and the loadings of the variables on the same plane. Variable selection, standardizing, sign flipping and the choice of component axes are all interactive. Output reproduces R's prcomp().

Open app GitHub

Classification tree (CART)

Build a CART classification tree and view decision rules, a confusion matrix, and variable importance.

Open app GitHub

Multiple linear regression

Fits a multiple linear regression model by ordinary least squares (OLS). Categorical (text) features are auto-detected and can be dummy-coded. Output is formatted like R's summary(lm(...)).

Open app GitHub

Logistic regression

Fits a logistic regression by maximum likelihood (IRWLS / Fisher scoring). The response need not be binary: pick any categorical column and choose which value counts as 1, and every other value becomes 0. Also reports odds ratios, an ROC curve and a confusion matrix. Output reproduces R's summary(glm(..., family="binomial")).

Open app GitHub

Cox proportional hazards model (survival analysis)

Fits a Cox regression from a follow-up time and a censoring indicator instead of a single response. Ties use the Efron approximation (R's default) and the partial likelihood is maximised by Newton-Raphson. Reports hazard ratios with 95% confidence intervals, the concordance, and the likelihood-ratio, Wald and score (logrank) tests, plus Kaplan-Meier, log-log and Schoenfeld-residual plots for checking proportional hazards. Output reproduces R's summary(coxph(...)).

Open app GitHub

R Shiny apps (webR)

R Shiny apps that run entirely in your browser with webR (R compiled to WebAssembly). No server is involved, and the data you load stay on your computer. The first launch takes a while because R itself is downloaded.

Regression & Classification Tree (CART)

Builds regression and classification trees (CART) with R's rpart. Shows tree diagrams (rpart.plot, partykit), the ROC curve, variable importance and the complexity-parameter plot, and supports a train/test split. Figures can be saved as PDF/PNG.

Open app GitHub