Free Error Bar Chart Maker

Error bars answer one of two different questions, and a reader cannot tell which from the picture alone. Standard deviation bars answer "how much do the individual measurements vary?" Standard error and confidence-interval bars answer "how precisely do we know the mean?" A chart that does not say which it is cannot be read.

Pick SD when the spread itself matters: how consistent plant growth was, how varied reaction times were. Pick a 95% confidence interval when the point is comparing means, because it has a direct interpretation and grows honestly wider when a group is small. SE bars are common in older papers and some fields, but they look reassuringly short, so state them clearly.

Whichever you choose, GraphMake writes it under the chart ("Error bars show ±1 standard deviation", "95% confidence intervals") and adds the group sizes when it did the maths, so the exported figure carries its own legend.

Paste every measurement and the studio does the arithmetic a spreadsheet would. The mean is the average of each group. The standard deviation uses n − 1 (the sample standard deviation, what Excel calls STDEV.S). The standard error is the SD divided by the square root of n.

The 95% confidence interval is the mean ± t × SE, where t comes from Student's t distribution with n − 1 degrees of freedom: 2.776 for 5 values per group, 2.262 for 10, 2.093 for 20, approaching 1.96 for very large groups. Small groups therefore get honestly wider intervals, which is exactly what a rounded 1.96 would hide.

Groups can come as rows (a label, then its values) or as spreadsheet columns under a header row, with blank cells allowed, so most exports paste in without rearranging.

Overlapping error bars do not prove two groups are the same, and separate bars do not prove a difference. The rule of thumb that holds up best: when two 95% confidence intervals do not overlap at all, the difference between those means is statistically significant. For anything closer, run the proper test and report it alongside the chart.

Keep bars starting at zero. A bar's length is its value, and a cut axis exaggerates small differences, which is why this maker only lets points, not bars, use a tighter axis.

Label the groups and the unit on the axis, and keep the caption. Those three things are what separate a figure a reader can check from a picture they have to trust.