Bar Graphs

by Juan C. DÃ¼rsteler

[message nº 157]

Bar graphs or bar charts are ubiquitous visual representations, but they are not always correctly used and many times we donâ€™t get the most out of them. We review here their essence, properties and utilization. 
Who hasnâ€™t ever used a bar chart to support a presentation or to study how sales have developed in the last quarter? We all appear to be very familiar with this type of graphic. Nevertheless
many times we fail to notice some of its possibilities and limitations.
What is it?


Bar graph 
Column graph 
Generically
called "Bar Graphs" 
A bar graph is that two dimensional graphic representation where the elementary graphic objects are a set of rectangles drawn in parallel so that the extension of the same is proportional to the magnitude they intend to represent.
The rectangles, or bars, can be either horizontally or vertically positioned. In the latter case they receive the name of
column graphs or charts.
In the following we'll refer to both of them as "bar
graphs" independently of the orientation of the rectangles.
Use.
Typically they are used to
 compare magnitudes among several categories or
 the evolution in time (the change) of a particular
magnitude.
 the comparison of the evolution in time of several
categories, i.e., they are also used as a blend of the two
preceding uses.
Types of Data
As we know data can be classified into three types:
 Categorical or Nominal (birds, mammals, tall, small, green or blueâ€¦)
 Ordinal or Sequential (the months of the year, the alphabetâ€¦)
 Numerical or Quantitative (everything that can be represented with numbers)
The data variables mainly used in bar graphs are those of categorical and
ordinal types (especially the temporal ones).
This type of representation is not very suitable for representing quantitative data.
Components
A bar graph consists, at least of:
Components of a bar or column graph 



Quantitative Axis
A quantitative axis with a linear scale that serves as a reference for the magnitude of the variable to be represented. In a column graph this is the ordinate axis
[Y] and in a bar graph itâ€™s the abscissa axis [X]. This axis can hold negative values. 
Categorical
or Ordinal Axis
A categorical or sequential axis where the categories or elements of the sequence are placed (the abscissa axis
[X] in a column graph or the ordinate axis [Y]
in a bar graph). This axis is perpendicular to the quantitative one. 
A
Set of Rectangles
A set of rectangles whose extension parallel to the quantitative axis is proportional to the magnitude of the category or sequence represented along the axis. 
Main types of bar graphs
There are many types and variants of the same. We will discuss here the
most relevant ones.
 Simple
It contains only one data series (for example sales of the same country along different months).
 Grouped
It contains several data series, for example the monthly sales in several countries. In this case the sequential axis would contain the months and the quantitative one the sales figures. Each data series is represented as a set of rectangles that share colour or texture.
In each category or sequence the rectangles are usually grouped together while space lies between groups and categories or sequences. For example in the sales case, the obtained results by France, UK and Italy in the month of November would be represented as three rectangles with heights proportional to their value, placed side by side in the space allocated in the sequential axis for the month of November.
 Overlapped.
Itâ€™s a bar graph where the elements of the group partially
overlap instead of laying side by side. If the overlapping is 100% they can be easily confused with the stacked bar graphs.



Simple 
Grouped 
Overlapped 

Stacked
Also called segmented or extended, itâ€™s similar to the grouped one but in this one each of the segments in which the bar or column is divided belongs to a different data series.
It shows how a total entity is subdivided into parts. If the non quantitative axis is temporal it allows you to show how these proportions or compositions vary with time. For example you can make a stacked bar graph with the monthly sales in each country one on top of the other in such a way that the total height of the bar depicts the total amount of sales and the proportions of the segments give you an idea of their participation in the whole.
 Linked or connected
If you add lines connecting the places of each bar where thereâ€™s a change of segment it receives the name of linked or connected bar graph
.
 Hundred per cent.
A stacked bar graph where the total extension completely covers the quantitative axis
so that what the segments show is the percentage of them into the total, that represents 100%.



Stacked 
Linked or Connected 
100% Stacked 
 Floating, bidirectional or paired..
In this case the line of 0 value acts a separator of two bar graphs that share this line as a base line but where each of them extend their bars in opposite directions. A typical case are the graphics of the demographic pyramid, where to the right you find the amount of surviving males of a certain age and to the left the equivalent quantity of females.
 Pictorial..
So called when the bars are constituted by the repetition of a series of symbols typically showing the nature of the data. For example, a series of motorcycles or cars, one after another, to depict the comparative sales of bikes and cars. The symbols can be of equal size or can be distorted to fit the required length of the bar.
 Range
In this type of graphics the maximum and minimum extension of the bars indicate the upper and lower boundaries of the data values. Sometimes internal values are designated by showing a line crossing the
bar that can depict a statistical magnitude like the average, etc.



Paired or Bidirectional 
Pictorial 
Ranges 
There are many variations and combinations of the above mentioned types that we donâ€™t have space to describe. In the next issue weâ€™ll review the best practices to use when building appropriate bar graphs.
Commendable books on the topic
Links of this issue:
