I completely agree that the bar chart is completely useless method of presenting more than very few series. However for data like your example, I prefer an actual table with the background of each cell colored using the same kind of color scale you did in the bubble chart. The background colors effectively turn the table into a heatmap which very quickly can give an overview of the “shape” of the data. Then the literal numbers offer higher accuracy for drilling in. In comparison to the bubble example the color filling the whole cell rather than just inside the bubble makes it easier to see the overall pattern (like a low resolution picture with large pixels), without being broken up by the repeating circles and large amount of unused white space in the bubble chart. The bubble size only gives an approximate impression of magnitude of the value, but this isn’t that helpful as color scale already does that and does it better as you can show things like negative values with contrasting colors that can’t be mapped to circle size, and while human visual systems is adept at picking our shapes in a color field, it’s not really adapted to surveying a large number of different objects of different sizes and picking out patterns from that.
Thanks! Love your post! They are very helpful! I need an advice, I’m struggling with charts (I’m only using power bi) so I need a way to start creating charts
Thanks for sharing this. It functions effectively if you're less concerned with the precise numerical value and more focused on gaining a qualitative understanding. Otherwise, as stated in another comment, I also would prefer an heatmap.
I completely agree that the bar chart is completely useless method of presenting more than very few series. However for data like your example, I prefer an actual table with the background of each cell colored using the same kind of color scale you did in the bubble chart. The background colors effectively turn the table into a heatmap which very quickly can give an overview of the “shape” of the data. Then the literal numbers offer higher accuracy for drilling in. In comparison to the bubble example the color filling the whole cell rather than just inside the bubble makes it easier to see the overall pattern (like a low resolution picture with large pixels), without being broken up by the repeating circles and large amount of unused white space in the bubble chart. The bubble size only gives an approximate impression of magnitude of the value, but this isn’t that helpful as color scale already does that and does it better as you can show things like negative values with contrasting colors that can’t be mapped to circle size, and while human visual systems is adept at picking our shapes in a color field, it’s not really adapted to surveying a large number of different objects of different sizes and picking out patterns from that.
That great!
Was struggle to make in order Xaxes with bar and scatter made my day!
Thx, Bro!
Thanks! Love your post! They are very helpful! I need an advice, I’m struggling with charts (I’m only using power bi) so I need a way to start creating charts
It's a mistake to use a bar plot when your x axis is a time series. A line plot is almost always better
If it's important to know the exact values in the chart, a heatmap is another good choice. https://seaborn.pydata.org/generated/seaborn.heatmap.html
Thanks for sharing this. It functions effectively if you're less concerned with the precise numerical value and more focused on gaining a qualitative understanding. Otherwise, as stated in another comment, I also would prefer an heatmap.