Nice going on the comparison of the pie with other chart types! I chose the scatterplot because it makes for immediate decoding of quantities on both axes (rather than reading labels or assessing surface area of pies and wedges). It also retains the quantitative info of the total revenues (lost in the bar graph), but without the clutter (and the indirect readout due stacking of the bars) on the y-axis of the Manderley. Also, it is the only graph that reveals the relationship between the two variables: the lower the total revenue, the higher the proportion of arms revenue. Finally, it is the best in terms of data ink/non-data ink (signal to noise ratio).
Minor: the title of that graph should read 'billions of dollars' because that's what the y-axis depicts.
Hi Jon (and Maarten): totally +1 re Maarten's comments. For me, it's no question that the scatterplot is more immediately meaningful and useful - as Maarten says, it's both efficient, and it provides richer insight with the additional value dimension.
Conversely, the pie charts are greedy in real estate terms, and they impose a greater cognitive load - the reader has to re-impute the scale in their mind for each pie because they're not on the same plot area.
The bar chart is better than the pies for making comparisons but it suffers from being limited to just the one value dimension.
I *want* to like a Marimekko, but I've not seen a successful one in the wild as yet. They're just too hard for making sense of relativities across two dimensions - we can't hope to gauge accurately for instance whether a longer, slimmer green bar represents greater volume than a shorter, taller bar. Maybe a Marimekko works best with much fewer data points...
Hi Jon,
Nice going on the comparison of the pie with other chart types! I chose the scatterplot because it makes for immediate decoding of quantities on both axes (rather than reading labels or assessing surface area of pies and wedges). It also retains the quantitative info of the total revenues (lost in the bar graph), but without the clutter (and the indirect readout due stacking of the bars) on the y-axis of the Manderley. Also, it is the only graph that reveals the relationship between the two variables: the lower the total revenue, the higher the proportion of arms revenue. Finally, it is the best in terms of data ink/non-data ink (signal to noise ratio).
Minor: the title of that graph should read 'billions of dollars' because that's what the y-axis depicts.
Maarten Boers
Hi Jon (and Maarten): totally +1 re Maarten's comments. For me, it's no question that the scatterplot is more immediately meaningful and useful - as Maarten says, it's both efficient, and it provides richer insight with the additional value dimension.
Conversely, the pie charts are greedy in real estate terms, and they impose a greater cognitive load - the reader has to re-impute the scale in their mind for each pie because they're not on the same plot area.
The bar chart is better than the pies for making comparisons but it suffers from being limited to just the one value dimension.
I *want* to like a Marimekko, but I've not seen a successful one in the wild as yet. They're just too hard for making sense of relativities across two dimensions - we can't hope to gauge accurately for instance whether a longer, slimmer green bar represents greater volume than a shorter, taller bar. Maybe a Marimekko works best with much fewer data points...
So as I say, scatterplot for me, no contest!