Not really, if you are looking at variations around e.g. Labour votes around 10 million, then starting y-axis at zero would make the chart pretty well useless, starting at the lowest point in the time series is the approach that makes sense.
I respectfully disagree. A bar chart implicitly shows values proportionate to the bar’s length. If you want to do what you did to show the variation better I believe the recommended way is to include a zigzag break across the bottom section of the bars to show you’ve truncated them. Stephen Few. Edward Tufte have written books on good data presentation. Contentious to some :-)
Minor irritation - bar graphs starting at non-zero… straight out of “how to lie with statistics”. Otherwise quite an interesting POV.
Not really, if you are looking at variations around e.g. Labour votes around 10 million, then starting y-axis at zero would make the chart pretty well useless, starting at the lowest point in the time series is the approach that makes sense.
I respectfully disagree. A bar chart implicitly shows values proportionate to the bar’s length. If you want to do what you did to show the variation better I believe the recommended way is to include a zigzag break across the bottom section of the bars to show you’ve truncated them. Stephen Few. Edward Tufte have written books on good data presentation. Contentious to some :-)
OK understood.