Robert Kosara: data visualization as a research, teacher and practinionner
course description and notebook set
march 7th, youtube
- mapping each object at position on x and y
- range = axis
- general view / overall shape: relationship/correlations in the dataset
- things that stick out : unusual values (function differently in average)
plot title => tooltip!
anscombe dataset different views (to check again)
- sort
plot sort: {x: "y", reverse: "true", limit: 8}
-
compare values
-
bar charts vs tables
- tables are great for precise data
- visualisation is for structure and patterns. precision is not usually what is wanted
- If you need precise numbers, either use a table or label/tooltip
-
curtting off the x axis to remove 0-300 and keep only 300 to 600
- theoretically one can see the differences better
- however, we look at bars as length
- when you cut off the axis, you don't realise the biais
- misguiding: people might not check the
- the point of the chart is to show the numbers to understand them, and intuitively understand the difference, which you can't do if you need to remap the values
cardinal sin of data visualisation
- too similar to bar charts, people mistakenly use pie or bar charts for the wrong reason
- distribution of part to a whole not good at comparison between sizes
- categorical axis: color
- number: length on the arc
- very hard to guess the differences between two areas
- donut chart (whole in the center)
- often considered a bad idea
- will develop next time
- standard line chart: only one possible value for the position on the x axis
- assume that there is a continuity: draw lines between the positions
- common rule (though not proven by research): rough overall 45° "banking on"
Plot.lineY
- radar charts: very dependent on the order of the radial axis (especially the area can change drastically depending on position)→ rather poor chart