How choose the number of classes of a series? I am newbie in DS.
I'm doing a choropleth map with votes of a senator in several cities.
And I have two questions:
1 - How choose the correct number of classes to part the data?
At the time im using four classes, but not for a specific reason.
2 - How choose the technique to use to create these partitions? I saw there are some like natural breaks, quantiles, standart deviation etc, but I don't know how and why to use.
The data looks like this.




city
votes




city_1
1k


city_1
23k


city_1
57k


city_1
112k


.



.



.



city_n
500k




Could you point me the theory around this?
Thank you all.
 A: It sounds like you might be interested in the literature on cartography and, more generally, data visualization.  The former explicitly addresses questions of choropleth (area-filled) maps.
Among the considerations in any such work are

*

*Purpose: Why do you want to draw this map?  What message should it convey?


*Audience: To whom do you wish to communicate?


*Psychology: What are the ways in which maps and statistical graphics communicate information?

*

*How do people derive meaning from graphics?

*What methods of symbolization lend themselves to faster and more accurate understanding?



*Statistics: To what extent do the data suggest forms of visualization and grouping of the values?  What are suitable ways of expressing and re-expressing data for visualization?
The classic reference is Jacques Bertin, Semiology of Graphics.  It contains extensive (thousands) of examples organized by a few basic principles.
A modern study of the psychological aspects is Allen MacEachren, How Maps Work.  It is difficult reading but provides rewarding insight.
Another classic work is William Cleveland, Visualizing Data.  Cleveland's group at Bell Labs conducted psychological studies aimed at obtaining scientific answers to these basic questions, leading to practical advice for constructing statistical graphics (and, incidentally, maps).
