See the book "Exploratory Analysis of Spatial and Temporal Data: A Systematic Approach" by Natalia Andrienko, Gennady Andrienko, especially 4.2 "The value of visualization".

http://books.google.de/books?id=Oqq7oP31EycC&pg=PA166&lpg=PA166&dq=%22Exploratory+Data+Analysis%22+and+%22Information+Visualisation%22&source=bl&ots=qEYOz2ZnfY&sig=Qfapns1aYdSK2O4PonkBeUQVCzk&hl=en&sa=X&ei=q4jXU-jaD8eGywPW2YKABQ&redir_esc=y#v=onepage&q&f=false

Thank you very much!

The book is very interesting, I'm trying to make sense of it all crossing this new information with what I read previously. Would you agree to the following?

1. EDA is a process that doesn't necessarily answer a question but rather "explore" the data available.

2. IV is about statistcs and visualising the data and can be seen as a "tool" for EDA

3. EDA goes further than IV as it is also looking at clustering, classification, association... that are Data Mining (DM) techniques.

In conclusion, relationship between EDA, IV and DM:

- IV is limited to statistics and visualisation and could be opposed to DM that look further into the data, trying to find patterns and infer new information.

- EDA is a more general term about the use of the data (exploring it) and could, for instance, use IV to show DM output.

Thank you for any feedback you could give me on my understanding!

.

my two (rough) words : EDA is about building a good representation space from the data (therefore tackling the problem of including in this representation as much information as possible without overfitting) ; IV relies on this representation to synthetize some aspects of this representation

EDA without IV does not help much in solving complex problems

IV without EDA is the quickest way to results with no statistical significance

.

As far as I got it, DM is nothing else but EDA, only applied to electronic data bases (DM as the analysis step in the process of "Knowledge Discovery in Databases", KDD).

IV is a (actually the most important) tool of EDA.

In principle, yes.

A small add. EDA is much more then "explore" data, and it can answer a lot of question. IV is a helping tool, yes, but limited in its applicability. It helps to have a first look at the data (think of it as a first layer), but to obtained a more detailed pictures (2-nd, 3-th layer, and so on) you need more than only visualizing the data.

Fantastic!

Thank you so much to all of you, everything is much clearer now.

for me 

EDA is when I don;t understand my data (most of the time) and I just plot everything against everything else.

IV is used when I want to explain science in understandable ways.

Thanks, 

I understand the EDA part but I'm not about what you mean by "explain science in understanble ways" for IV. 

Could you elaborate?

right

I guess  I didn't explain that in an understandable way :)

Visualization is a very powerful way to potentially explain complex science to a variety of different audiences (students, colleagues, lay people) and the power of visulization to accelerate understanding should never be underestimated.  On the other hand, its  quite easy to make poor visualizations which can be misleading.

Recently (within the last year) a significant number of open source Viz/chart  libraries are now available

here are two examples:

www.highcharts.com

https://github.com/mbostock/d3/wiki/Gallery

In my areas of research (climate change, astrophysics, renewable energy) all of my data (as well as the bulk of most data) is N-dimensional in the sense that there is a geometric coordinate point (lat, long) with multiple attributes at the point.  The power inherent in IV is to be able to represent that, in understandable ways, using just one graphical interface.

Okay,

That is now explain in a very very understandable way ;)

Thank you very much for your interesting explanation, and the link to very useful libraries.

I am feeling much more confident about it all.

Having no previous knowledge of IV, I was curious to find out what Wikipedia has to say about it and also about EDA and DM. I don't think this thread brings much clarity to the original question. I found Wikipedia much more helpful.

Yaelle, I think the idea about EDA and IV that you presented in your original question is correct. However, in my humble opinion, some of the conclusions you drew later on in this thread are incorrect.

I am referring to your points #1 and #2. They appear quite incorrect to me, since EDA is all about statistics and has specific objectives it attempts to answer. Whereas from Wikipedia I get the idea that IV is about looking for creative ways to visualize data in order to learn something new in a much more informal way. DM also has a specific goal and statistical and other methods are used to achieve it.

I think your question is excellent. By no means, do I claim to be correct as I, too, am just trying to make sense of all this.

I recommend reading Wikipedia. The IV article makes a very helpful distinction between scientific visualization and information visualization.

http://en.wikipedia.org/wiki/Information_visualization

http://en.wikipedia.org/wiki/Exploratory_data_analysis

http://en.wikipedia.org/wiki/Data_mining

Natture methods, 2014.

Hello,

Perhaps it is a bit late to reply but I found this question really interesting so here it is my attempt.

I'd say that EDA is the process of gaining relevant knowledge about a complex system by exploring data and IV is the process of "telling what you learnt", that is, showing these knowledge in a visual understandable friendly (and possibly interactive) way.

In conclusion, I guess they are complementary concepts that apply radically different tools, skills and approaches but that are perfect partners in the context of data analytics.

BTW: I enclose a link a to related question about the differences between "exploratory data analysis" and "confirmatory data analysis". I hope you find it interesting...

What do you think?

http://www.butlerscientifics.com/#!Exploratory-vs-Confirmatory-Research/ciyl/CA5AF661-7462-4AE8-815C-04B6F70BCC76