An error bar is a line going through a point. This line shows the variation of the data of that point. Actually, these bars show standard deviation. If the data for a point is in concentration to the mean, then the line will be small, showing low standard deviation and vice-versa.
Why do you need it?
We use Error bars with other visualization to show their accuracy. The length of the error bar is directly proportional to the variation of data around a mean value. The error bar is always parallel to the measure-axis. This all helps us to analyze our data more accurately and effectively.
What kind of data you can visualize with it?
Assume you have a bar graph showing a measure of the population on the Y-axis and age-bracket on the Y-axis, then error bars are really helpful to show the accuracy of the dataset. Some other examples on which error bars can work are Area chart, histogram, line graph, multi-set bar chart, scatterplot, etc.