I’m not sure what point it’s trying to make. If it’s trying to make a relation between urban population and transit share those should be the x and y axes and it should be a scatter plot. I don’t know if time is even relevant to that (you could average over time, or present separate plots for the time scale).

Or if it’s meant to compare the trends over time, they should be either separate graphs or the axes should be such that the lines of the two groups don’t overlap significantly. You can do dual y-axes and have a graph make sense, but only if your scaling is sensible. Although considering how many cities/colors are being used it’s questionable if that approach would work here. I’m definitely not wild about the solid vs. dashed method to try to specify which data set is being represented either (again, it can work when done properly but does not here).

Or maybe even just break it down by city. It’s trying to do far too many things in a single graph and not doing it well.

-Mathematician/data analyst with a graph fetish and past experience in scientific research.

Wayne

Oh, actually regarding the first idea, you could drop the distinction between cities if you don’t care about differences based on particular area and just use symbol/color coding based on time to represent any change over time in the scatter plot.

Thomas MercierDo you mean other than that it should actually be two graphs since the use of the two vertical axis makes for a difficult comparison?

Bill LindekeModerator Post authoryeah it hurts my head a bit like if you watch a 3D movie without those glasses on

David MarkleI can’t make sense of it, with two vertical scales and dotted versus solid lines.

Nathanael1980 is a weird place to start the graph. There was a temporary boom in ridership due to the Second Oil Crisis of 1979.

It also violates Tufte’s rules of clear graph presentation, as David Markle points out.

Pingback: Sunday Summary – December 21, 2014 | streets.mn

WayneI’m not sure what point it’s trying to make. If it’s trying to make a relation between urban population and transit share those should be the x and y axes and it should be a scatter plot. I don’t know if time is even relevant to that (you could average over time, or present separate plots for the time scale).

Or if it’s meant to compare the trends over time, they should be either separate graphs or the axes should be such that the lines of the two groups don’t overlap significantly. You can do dual y-axes and have a graph make sense, but only if your scaling is sensible. Although considering how many cities/colors are being used it’s questionable if that approach would work here. I’m definitely not wild about the solid vs. dashed method to try to specify which data set is being represented either (again, it can work when done properly but does not here).

Or maybe even just break it down by city. It’s trying to do far too many things in a single graph and not doing it well.

-Mathematician/data analyst with a graph fetish and past experience in scientific research.

WayneOh, actually regarding the first idea, you could drop the distinction between cities if you don’t care about differences based on particular area and just use symbol/color coding based on time to represent any change over time in the scatter plot.