Compare distribution shapes, quartiles, medians, means, and individual observations in one responsive workspace.
Compare the shape and center of travel times across different day types.
| Group | Count | Minimum | Q1 | Median | Mean | Q3 | Maximum | Std. dev. | Bandwidth |
|---|
A violin plot combines a smoothed density profile with compact statistical markers. Wider sections indicate values where observations are more concentrated; narrow sections indicate sparse regions. Unlike a box plot alone, the silhouette can reveal skew, long tails, shoulders, and multiple peaks.
Put each comparison group in its own numeric column and place observations down the rows. Import CSV, TSV, TXT, JSON, XLSX, or XLS data, then select the numeric columns to use. The editable table lets you correct values, rename groups, change colors, hide series, and add observations directly.
The violin outline is a Gaussian kernel density estimate rather than a literal histogram. Every observation contributes a smooth curve, and those curves are combined into the displayed shape. The bandwidth determines how strongly nearby observations blend together.
A smaller bandwidth reveals more local detail but can create unstable bumps in small samples. A larger bandwidth produces a smoother silhouette but may hide meaningful structure. The default uses Silverman's sample-based rule multiplied by the bandwidth control, making the calculation reproducible for the current values.
Equal maximum width gives every violin the same peak width and is useful for comparing shapes. Equal area keeps density areas comparable on one shared scale. Count scaling gives larger samples more visual weight. Always state the scaling rule when the apparent width could affect interpretation.
The inner box marks the first and third quartiles, while its central line marks the median. The mean uses a separate orange marker so it cannot be confused with the median. A visible separation between mean and median often accompanies skew, but the full silhouette and sample context should guide interpretation.
Individual observations help reveal sample size and the smoothing behind small violins. They can become visually dense for large datasets, so the display is optional. The computed statistics table remains available for exact counts, quartiles, range, mean, standard deviation, and effective bandwidth.
Density peaks are estimates, not directly observed categories. Small samples, repeated values, boundaries, and bandwidth choices can materially change the outline. Compare groups measured under consistent conditions and do not infer population structure solely from a decorative-looking bump.
Use vertical violins for short group names and horizontal violins when labels need more room. PNG and SVG exports include the visible groups, current density settings, statistical marks, legend, grid, titles, colors, and optional transparent background.
Place each class or teaching method in a separate sample column. The violin can reveal whether scores are concentrated near one level, spread broadly, skewed, or split into multiple clusters. Median, mean, and quartile marks add exact landmarks, while optional sample points show how much evidence supports the smoothed outline.
A wide section indicates that more probability density is estimated near that value; it is not a count printed directly from the data. The interpretation also depends on the scaling mode. Equal-width violins emphasize shape, while area or count scaling can retain information about relative sample size.
With few observations, the kernel can create a polished silhouette that looks more certain than the data justify. Show points, report group counts, and test nearby bandwidths. Use a box plot when a compact, less smoothing-dependent summary is preferable.
Groups should use the same units, collection process, eligibility rules, and period. A different shape may reflect measurement or sampling differences rather than a real population contrast. Treat apparent modes as hypotheses to investigate, not automatic proof of distinct subgroups.