Error Bar Chart Maker
Enter each group's mean, standard deviation and sample size — or paste the raw values — and get a chart of group means with error bars you choose: standard deviation, standard error or a confidence interval. The kind of bar is written on the image, a ready-made APA figure note says what the bars represent, and a table shows all three measures so readers can convert. Download as SVG or PNG. Free, no sign-up, and your data stays in your browser.
Write Control, 4.2, 1.1, 24 — the group's name, its mean, its sample standard deviation and its number of observations. A first line of column names is ignored. Use semicolons between cells if your numbers use decimal commas.
A CI is the range of means the data are compatible with. For independent groups of 10 or more, 95% CIs that overlap by about half of one arm correspond to p ≈ .05 (Cumming, Fidler & Vaux, 2007).
3 groups
Show every group's statistics as a table
| Group | n | Mean | SD | SE | 95% CI lower | 95% CI upper |
|---|---|---|---|---|---|---|
| Control | 24 | 4.2 | 1.1 | 0.22 | 3.74 | 4.66 |
| Low dose | 25 | 5.6 | 1.3 | 0.26 | 5.06 | 6.14 |
| High dose | 23 | 7.1 | 1.6 | 0.33 | 6.41 | 7.79 |
Suggested figure note
Error bars represent 95% confidence intervals. Group sizes: Control, n = 24; Low dose, n = 25; High dose, n = 23.
Title and figure style
Takes the title out of the image and sets the figure number in bold and the title in italics above it — the APA 7 layout.
How to Use the Error Bar Chart Maker
- 1. Enter your groups. With Summary statistics, write one group per line as
Name, mean, SD, n. With Raw values, writeName: value, value, …and the tool works out the mean and SD itself. - 2. Choose the error bar. Pick SD, SE or a confidence interval. For a confidence interval, set the level (95 is the usual one). The chart is labelled with the choice, and the hint below the buttons says what it means.
- 3. Choose columns or dots. Columns always start at zero, because a column's height is its mean. Dots fit the axis to the means and their bars, which suits means far from zero.
- 4. Add the figure note. Use as the APA figure note writes a sentence saying what the bars represent and how large each group is, and turns on the APA layout.
- 5. Download. Take the SVG or PNG. Copy statistics gives a table of n, mean, SD, SE and the interval limits for every group.
SD, SE and Confidence Intervals
The three bars answer different questions, and the chart is only honest if it says which one you drew. The standard deviation describes the spread of the data and does not shrink as the sample grows. The standard error of the mean, SD ÷ √n, describes how precisely the mean is estimated. A confidence interval is the mean ± t × SE, with t taken from the t distribution on n − 1 degrees of freedom; for samples of 10 or more, a 95% interval is close to mean ± 2 × SE.
| Bar | Half-length | Shows | Gets shorter with a bigger sample? |
|---|---|---|---|
| SD | SD | How spread out the values are | No |
| SE | SD ÷ √n | How precisely the mean is estimated | Yes |
| 95% CI | t × SD ÷ √n | The range of means the data are compatible with | Yes |
Worked example from the default data
Control group: mean 4.2, SD 1.1, n = 24. SE = 1.1 ÷ √24 ≈ 0.22. For a 95% interval, t on 23 degrees of freedom is about 2.07, so the half-length is 2.07 × 0.22 ≈ 0.46 and the interval runs from about 3.74 to 4.66. The default data are invented for illustration.
Reading Error Bars Honestly
Cumming, Fidler and Vaux (2007) set out rules of thumb that apply to independent groups of about 10 or more. If SE bars have a gap between them of about one SE, the difference is roughly at the p = .05 level; if 95% confidence intervals overlap by about half of one arm, it is again about p = .05. Overlapping bars do not by themselves show that two means are the same, and none of this applies to repeated measurements on the same participants, where a paired analysis is the right test. They also recommend stating in the legend what the bars are and how many observations each group has, which is what the suggested figure note does.
Cumming, G., Fidler, F., & Vaux, D. L. (2007). Error bars in experimental biology. The Journal of Cell Biology, 177(1), 7–11. https://doi.org/10.1083/jcb.200611141
- Say what the bars are. The chart prints it above the plot; say it in the figure note too.
- Give n. A bar means little without the group size behind it.
- With very small groups, show the points. With n around 3, a confidence interval is wide and an SD is unstable; showing every value tells more.
Frequently Asked Questions
- What is the difference between SD and SE error bars?
- SD bars show how spread out the individual values are. SE bars show how precisely the group mean is estimated — the SD divided by the square root of n — so they get shorter as groups get bigger. Use SD to describe your sample and SE or a confidence interval to talk about the mean.
- Should I use SD, SE or a 95% confidence interval?
- It depends on the question. SD describes the data. For comparing means, a confidence interval is usually the clearest, because it is directly interpretable; SE bars are about half as long and can look more impressive without saying more. Whichever you choose, label it.
- How do I calculate a 95% confidence interval from the mean, SD and n?
- Work out the standard error, SD ÷ √n, multiply it by the t value for n − 1 degrees of freedom at 95% (about 2.0 for large n, 2.26 for n = 10), and go that far above and below the mean. This tool does it for each group and lists the limits in the table.
- How do I describe error bars in an APA figure note?
- State what the bars represent, for example Error bars represent 95% confidence intervals., and give the group sizes. The tool writes that sentence for you; turn on APA style figure to place it as the note under the image.
- Why do the columns start at zero but the dots do not?
- A column's height encodes its mean, so a column that starts above zero exaggerates differences. A dot encodes only its position, so the axis can be fitted to the data. If your means are far from zero, dots show the differences more clearly.
- Why is there no error bar for one of my groups?
- A standard error or confidence interval needs at least two values, because it comes from the spread of the group. A group with one value has no spread to measure. The tool still plots its mean and says which groups have no bar.