please be significant
I'd used GraphPad Prism and knew how great it was, but I couldn't afford it; besides, my day-to-day data exploration never gets that complicated, and at heart I've always loved lightweight tools, so I simply built my own with AI.
Free statistics and scientific graphing in the browser. Paste a table from Excel, pick a test, and get the results, a methods paragraph and a publication-ready figure. Nothing is uploaded: everything runs on your machine, and the page keeps working offline once it has been opened.
Live: https://jieqianghe.github.io/plsig/
The interface can be switched between English and Chinese. Figures always stay in English, ready for papers.
| Data | Tests |
|---|---|
| One column | Descriptive statistics (CV, geometric mean, skewness, kurtosis) with Shapiro–Wilk normality; one-sample t-test and one-sample Wilcoxon against a hypothetical value |
| Two groups | Unpaired t-test (Student's or Welch's), Mann–Whitney U, paired t-test, Wilcoxon signed-rank |
| Three or more groups | One-way ANOVA with Tukey, Holm, Bonferroni or Dunnett post-hoc tests; Kruskal–Wallis with Dunn's test |
| Repeated measures (rows = subjects) | Repeated-measures ANOVA with Geisser–Greenhouse correction; Friedman with Dunn's test |
| Grouped (row titles × columns) | Two-way ANOVA (Type III sums of squares, so unbalanced designs work) with post-hoc comparisons; one-way ANOVA across every cell, with letters on each bar |
| XY | Pearson and Spearman correlation; linear regression with a 95% confidence band |
| Dose–response | Four-parameter logistic fit, EC50 / IC50 with standard errors |
| Survival | Kaplan–Meier curves, log-rank test |
| Contingency (counts) | Chi-square (Yates for 2×2), Fisher's exact test, relative risk, odds ratio |
Every statistic was checked against SciPy or pingouin. The Grubbs test flags possible outliers, but nothing is removed unless you exclude a value or a column yourself. A transform (% of the first column, log10, ln, average of replicates) can be applied before any analysis. Results re-run after every edit and come with a "Statistical analysis" paragraph ready to paste into a paper.
- Bar, scatter dot, box and violin plots with mean ± SD, SEM or 95% CI; XY scatter with fit; grouped bars; dose–response and survival curves; stacked contingency bars.
- Significance as asterisks, P values, brackets or compact letters. You can choose which comparisons appear.
- Click any text on the figure to rename it or make it bold, italic or a different size or colour. Click a bar to recolour it. Right-click selected bars to compare them in a new sheet.
- Style options: font size, dot size, bar width, hollow or hidden points, box or open axes, colour / hatched / grey fills, log axis, axis range, reference line, shaded band, broken y axis. Journal presets for Nature, Cell, Science and PNAS. Your favourite style can be saved as the default.
- Colour themes: Claude, Classic Prism, and NPG, AAAS, NEJM, Lancet, JAMA, JCO and D3 from ggsci.
- Export PNG (300 / 600 dpi) or SVG at journal column widths, or put every sheet into one panel figure lettered A, B, C. Copy to the clipboard as PNG or SVG. The SVG pastes into svguest for finer editing.
- The sheet behaves like Excel: select ranges; copy, cut and paste; fill down; insert, delete and drag columns; undo and redo.
- Paste from Excel, Google Sheets, Word, web pages or PDFs. Open .xlsx, .csv and .txt files; each worksheet becomes a tab.
- Several sheets at once, each with its own data, test and figure settings.
- Save a session as a
.jgzfile, or share it as a link that carries the data in the address itself. - Examples gallery: one ready-made data set for each kind of test.
index.html is the whole app: there is no build step and no server. Open it in a browser, or serve the folder with any static file server. sw.js makes it work offline. The only scripts it loads are jStat and the Plotly cartesian bundle, both pinned and integrity-checked from jsDelivr.
GPL-3.0. The ggsci palettes are GPL ≥ 3.