April 25, 2017
As someone who makes a lot of charts, I certainly agree that one shouldn't distort the data. So, if you had a column chart and one column was 90 units tall and the other was 100, you would want to display the y axis, starting from zero, to show how similar the columns were.
I did want to confirm the best practices for a line chart. In this case, I have a single line. It's data that I get every month, so the x axis will be up to 12 months. If the starting value of my line was 100, and it went up 10% over a year, then the ending value would be 110. If I were to display the y axis starting from zero, you really wouldn't see much of an increase. I can intensify the effect by zooming in on the y axis, to show more of an increase. The more I "zoom", the bigger the effect will be. So, it's a balance between not overselling a 10% increase, yet zooming in enough on the y axis, to show that something is going on i.e. a 10% upward trend.
Other than using artistic license, are there any best practices to displaying such line graphs with a single series? I realize a potential work around would be to just plot the increase from the 1st data point rather than the data itself. So, that graph would go from 0 to 10% on the y axis. However, I need to show the actual value, not the % increase in value.
Would appreciate input how others have handled this.
July 16, 2010
With line charts it's ok to start the vertical axis above zero as there are no columns to subconsciously compare the height of. The degree of effect you create is up to you. If you look at stock charts the scale is very small because often there is less than 1% change, whereas if you're plotting 10 points of change then you'd want a bigger scale.
Whatever you decide, I would label the points to reduce any effects caused by exaggeration.