# Raincloud plot maker

> Paste one column per group and get a raincloud plot: half violin, box plot and every data point. Summary table, PNG at 300 or 600 dpi. Free, no sign-up.

URL: https://www.labcanvas.com/tools/raincloud-plot-maker

Paste one column of values per group (names in the first row) and the tool draws a raincloud plot (Allen et al. 2019): for each group a half violin (the “cloud”, a Gaussian kernel density with Scott’s bandwidth, as SciPy), a box plot (median, quartiles, whiskers to 1.5 × IQR) and every value as a jittered point (the “rain”), horizontal or vertical. A summary table gives n, mean, SD, median, quartiles and the kernel bandwidth of each group. Download a 300 or 600 dpi PNG or open it in the LabCanvas Graph app, no code needed.

## Quick answer

- **What it does:** Draws each group’s kernel density as a half violin, a box plot and every data point, with a summary table.
- **Formula:** Cloud: f(x) = (1/nh) Σ φ((x − xᵢ)/h), h = n^(−1/5) × SD (Scott); box: Q1, median, Q3 at positions (n + 1)p, whiskers to 1.5 × IQR.
- **Example:** Control reaction times, n = 40, SD 103.2 ms: Scott’s bandwidth h = 40^(−1/5) × 103.2 = 49.3 ms; median 448.1 ms (Q1 424.3, Q3 532.1).

Free, in the browser, no account needed. Updated 2026-11-17.

## Features

- Half violin (kernel density), box plot and every value in one plot per group
- Gaussian kernel density with Scott’s or Silverman’s bandwidth, as SciPy’s gaussian_kde, cut at each group’s range
- Box plot: median and quartiles, whiskers to the last value within 1.5 × IQR
- Horizontal or vertical; clouds of the same width or the same area; box or points can be turned off
- Groups of different sizes; colour-blind-safe palettes (Paul Tol, Okabe–Ito)
- Summary table (n, mean, SD, Q1, median, Q3, IQR, bandwidth) to copy into Excel
- PNG at 300 or 600 dpi; opens in the LabCanvas Graph app, where tests can be added; nothing is uploaded

## How to use it

1. **Paste your groups**: One column per group with the names in the first row, straight from Excel. Groups can have different numbers of values.
2. **Choose the layout**: Horizontal (clouds above, rain below) or vertical; keep or drop the box and the points. Bandwidth, cloud size and colours are under More options.
3. **Export**: Download a PNG at 300 or 600 dpi, copy the chart or the summary table, or open the plot in the Graph app.

## At a glance

| | |
| --- | --- |
| Input | One column per group, names in the first row |
| Cloud | Gaussian kernel density, Scott’s or Silverman’s bandwidth (as SciPy), cut at the data range |
| Box | Median and quartiles ((n + 1)p rule), whiskers to 1.5 × IQR |
| Accuracy | Quartiles, means, SDs and bandwidths match NumPy and SciPy to 10⁻¹² |
| Export | PNG at 300 or 600 dpi, CSV summary; editable in the Graph app |

## How it works

**Why rainclouds.** A bar chart of means hides the data; a box plot hides the shape of the distribution; a violin hides the sample size. A raincloud plot shows all three at once: the density, the summary and the raw values (Allen et al. 2019), in line with the call to show the data rather than bars (Weissgerber et al. 2015).

**The cloud.** Each group’s density is a Gaussian kernel density estimate,

$$f(x) = \frac{1}{nh} \sum_{i=1}^{n} \varphi\!\left(\frac{x - x_i}{h}\right)$$

with the bandwidth \(h\) = factor × SD (sample SD, \(n - 1\)), the factor being Scott’s \(n^{-1/5}\) or Silverman’s \((3n/4)^{-1/5}\), exactly as SciPy’s `gaussian_kde`. The cloud is cut at the smallest and largest value of the group and drawn as a half violin, all clouds the same width (or the same area, so wide spreads look wide).

**The box.** The line is the median and the box runs from the first to the third quartile, computed at position \((n + 1)p\) in the sorted values (as GraphPad Prism; NumPy’s `method="weibull"`). Whiskers reach the last value within 1.5 × IQR of the box (Tukey).

**The rain.** Every value is a point, jittered sideways by a fixed pseudo-random amount, so the plot is the same every time it is drawn.

**Accuracy.** The plot is the LabCanvas Graph app’s raincloud graph, run in your browser. The quartiles, means, SDs and both bandwidths of the example match NumPy and SciPy to 10⁻¹² (scripts/raincloud-plot-maker-check.ts).

Sources: Allen M, Poggiali D, Whitaker K, Marshall TR, van Langen J, Kievit RA. Raincloud plots: a multi-platform tool for robust data visualization. *Wellcome Open Res* 2019;4:63, [doi:10.12688/wellcomeopenres.15191.2](https://doi.org/10.12688/wellcomeopenres.15191.2). Weissgerber TL, Milic NM, Winham SJ, Garovic VD. Beyond bar and line graphs: time for a new data presentation paradigm. *PLoS Biol* 2015;13:e1002128, [doi:10.1371/journal.pbio.1002128](https://doi.org/10.1371/journal.pbio.1002128). Scott DW. *Multivariate Density Estimation*. Wiley; 1992. Silverman BW. *Density Estimation for Statistics and Data Analysis*. Chapman & Hall; 1986.

## Worked example: reaction times in four groups

The built-in data are synthetic reaction times (ms) of 40 people in each of four groups, labelled as demo data. They are right-skewed, as reaction times are, and the alcohol group mixes two kinds of responders.

| Group | n | Mean | SD | Q1 | Median | Q3 | Bandwidth |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Control | 40 | 490.6 | 103.2 | 424.3 | 448.1 | 532.1 | 49.33 |
| Caffeine | 40 | 433.9 | 43.9 | 405.3 | 432.2 | 466.5 | 20.99 |
| Sleep-deprived | 40 | 574.6 | 100.8 | 504.5 | 551.4 | 626.4 | 48.19 |
| Alcohol | 40 | 559.4 | 86.7 | 486.3 | 544.2 | 627.4 | 41.45 |

**By hand (control).** Scott’s factor is \(40^{-1/5} = 0.478\), so \(h = 0.478 \times 103.2 = 49.3\) ms. Q1 sits at position \(41 \times 0.25 = 10.25\) in the sorted values, a quarter of the way from the 10th to the 11th value.

**What the plot shows that a box plot does not.** The alcohol group’s box looks much like the sleep-deprived group’s, but its cloud is broad and flat-topped, spread evenly from about 450 to 650 ms: the data mix people barely affected with people strongly affected (they were simulated as two kinds of responders), which a box alone cannot show. The control group’s mean (490.6) sits well above its median (448.1) because of a long right tail, which the cloud and the rain make plain.



## Questions

### How do I make a raincloud plot without code?

Paste one column per group (names in the first row) into the table. The plot appears at once; choose horizontal or vertical, then download a 300 or 600 dpi PNG or open it in the LabCanvas Graph app. No R or Python needed.

### What is a raincloud plot?

A plot that combines a half violin (the distribution’s density, the cloud), a box plot and the raw data points (the rain) for each group, introduced by Allen et al. (2019) to show the data, its summary and its shape together.

### Raincloud, violin or box plot: which should I use?

A raincloud when you want readers to see every value and the shape of the distribution; it works best with about 10 or more values per group. With very few values, a dot plot with the mean is clearer. A box plot alone hides bimodal groups.

### How is the cloud (density) calculated?

As a Gaussian kernel density with Scott’s bandwidth, n^(−1/5) × SD, the default of SciPy’s gaussian_kde; Silverman’s rule is under More options. The cloud is cut at each group’s smallest and largest value.

### Can I add statistics?

Open the plot in the LabCanvas Graph app and add a test (t test, ANOVA, Kruskal–Wallis and more); for p-value brackets on a box or violin plot, use the box plot maker.

### Can groups have different sizes?

Yes. Leave the extra cells of the shorter columns empty.
