What fraction of the standard deviation should the error bars be?

Question

I am certainly new to statistics. I did some simulations and got a lot of data. From the data I ran a AWK script to calculate the average $\bar x$; minimum, $x_0$ and standard deviation, $\sigma$ (the one where you divide by $N$, not $N-1$).

Now I want to plot the data. I guess, I can draw the histogram $\bar x$ high but I am confused how long my error bar should be, like should it be,

• one standard deviation long (68% confidence)
• or $2\sigma$ (95% confidence) or $3\sigma$ (99.7% confidence) long.
• or should I draw it from min-value to max-value

Answer 1

Error bars often represent one standard deviation of uncertainty, one standard error, or a particular confidence interval (e.g., a 95% interval). -Wikipedia

When you are talking about plotting standard deviation, are you sure that the data being plotted is not skewed? If it is, be careful enough to calculate true standard deviations before plotting them.

Answer 2

Consistent with the other comments, there is no universally agreed practice across all fields and disciplines. Thus, the length of your error bars should be:

• the standard as typically used in your industry / field;
• what you believe your audience ( or reviewers for journal publications!) most likely expect; and
• whatever you explicitly describe in the caption of your chart.

This last point means that the best way to avoid misinterpretation is to explicitly say in your chart what the length of the error bars represent!