Error bars. Whenever I see a plot without them, I think someone is trying to pull a fast one. Some scientific fields (or quasi-scientific) fields use them much

Error bars. Whenever I see a plot without them, I think someone is trying to pull a fast one. Some scientific fields (or quasi-scientific) fields use them much more than others. Quick-and-dirty physicists' method of generating an error bar when dealing with number of events in an interval is to assume that Poisson counting statistics apply, and take the square root of the number of events. Example: Two months ago we sold 400 units. Last month we sold 410 units. Should we celebrate? Square root of 400 is 20. The change in sales is within the margin of error. Celebration would be premature.