A chart offers the reader an interpretation of the data in a way that should be correct, clear and pleasant. Correctness is of course the most important thing, but aesthetic clarity and pleasantness are closely connected in presenting data correctly and effectively.

Graphs communicate first with shapes and colours. Relevant information passes through visual aspects, by using codes that are widely implicit and used unknowingly.

In fact, we choose to represent the data in a graphic form for its visual impact and capacity to synthesize the information, compared to a table or a written text. And this is the precise reason why, if the graphic is badly designed, the message will be misleading.

‘You see, but you do not observe.’ (A. Conan Doyle)

In particular, if we address an audience of non-specialists, we have to assume that the “normal” reader will not pay particular attention to the texts (except for the captions) (Swires-Hennessy 2014) and will interpret even the scale of the representation on the basis of his knowledge and conventions, for example by ignoring the annotations of the axes.

The cartesian coordinate system is more familiar and intuitive than the polar one (used in pie charts) and the same colours have their own signification system: the scale of intensity, for example, is intuitively associated with that of the values, so that light colours mean “less” and dark ones “more”.

When we perceive differences the world ‘takes form’ in front of us and for us (A. J. Greimas)

Another point to consider is that, in a sign system, meanings are constituted by differences. In a bar chart, for example, the reader will have to notice only the different lengths of the bars, without being “distracted” or “mislead” by other signs, like colours.

Monotony is “boring” and what breaks it, awakens the attention: a change in colours or style activates the reader’s interest, but it could be distracting from the focus of the message and make it more difficult to interpret – unless it is actually significant, and therefore necessary.

Less is more (L. Mies van der Rohe)

This is the reason why the most recent trends in this field go towards the radical simplification, according to the principle of “less is more“: few significant elements, to communicate in a clear and unambiguous way what really matters.

References

Chen, C., Hardle, W., & Unwin, A. (A c. di). (2007). Handbook of Data Visualization. Berlin: Springer Verlag.

Knaflic, C. N. (2015). Storytelling With Data: A Data Visualization Guide for Business Professionals (1 edition). Hoboken, New Jersey: John Wiley & Sons Inc.

Swires-Hennessy, E. (2014). Presenting Data: How to Communicate Your Message Effectively (1 edtion). Chichester, West Sussex ; Hoboken, NJ: John Wiley & Sons Inc.

Translated and adapted from “Ricerca sociale con R” (chapter 5 “I grafici”)
Fotografia: Library of Congress Catalog: http://lccn.loc.gov/20056768367

In 2018, R turned 25, and has become one of the most used languages for data analysis, probably the most required in the job market of data science, along with SAS and Python. It is an important skill for a dynamic and evolving job market, in a language increasingly adopted in the academic research community.

Not surprisingly, many are those who wish to approach this programming language, of which the difficulty is largely legendary.

Learning data analysis with R

Actually, the major challenges of learning R have to be faced at the beginning; afterwards, it becomes quite easy, even with more complex analysis techniques.

Do the difficult things while they are easy (Lao Tzu)

This characteristic is a good reason in itself to use R since the very first approaches with data analysis: being a language designed for statistics, it is easy to introduce the basic concepts of both, in parallel and with the necessary graduality.

# vector, object
V1 <- c(1, 0, 9, 2, 5, 3, 10)    
# classes
class(V1)                        
# calcolo della media
# base functions; length of a vector;
sum(V1) / length(V1)             
# mean function
mean(V1)                         
# next episode: missing values ...

This method, also followed in the text, can be daunting, especially if one does not intend to become a programmer or an experienced user, or is in a hurry to get a result.

Learning by doing with RCommander

In 2012, I decided to adopt the graphical interface of  RCommander in the training workshops on R, remembering when I had to perform a multiple correspondence analysis … without having ever used R before.

I hear and I forget. I see and I remember. I do and I understand (Confucius)

RCommander offers a “mixed” interface: the familiar drop-down menus, together with a window for the script, where all the activities executed from the menu are recorded. This allows the user to approach R procedures and functions without having to learn coding first, and also progressively learning the language. The reader will be accompanied in observing and studying the functions, according to the logic of learning by doing: the objective is to carry out an analysis of data, but, reflecting on procedures and results, one learns the structure of the language.

It is a didactic approach that works for the simple reason that almost everyone approaches software to use it, not to study it. As a matter of fact, R is used by researchers and professionals in sectors other than programming: economists, sociologists, psychologists, and scientists in all the fields of natural sciences. Nobody, in short, who is or wants to become a programmer.

The major risk in using a menu interface is to keep reliant on the interface, that is, limited in the possibilities of an andvanced use of R. In any case, the need to learn at least to read and modify the code emerges fairly soon. RCommander has several plugins (add-ons) for many procedures, but some have limited functionality. From the beginning, for example, the limits of the graphs produced are particularly evident, compared to the possibilities of the graphic environment of R.

To use R at a more advanced level, it is important to start with knowing the objects (data structures), and reading and writing the code, to get to work in what I call the mode “data + code”, that is the normal mode, more efficient and reproducible.

I’m not a great programmer; I’m just a good programmer with great habits. (Kent Beck)

To initiate the reader into the practice of working with scripts, from the second to the fourth chapter RCommander will be used to produce three reproducible scripts: one for importing and transforming data; one for descriptive analysis; the third for bivariate analysis and hypothesis testing.In parallel, the functions used are unfolded by means of simple exercises of code writing. The reader can therefore choose whether to work exclusively with RCommander, or pass directly or gradually to the language of R.

Translated and adapted from “Ricerca sociale con R” (“Introduction”, pp. 7-9)
Ricerca sociale con R: (Italian)