A few months ago, I was reviewing email from one of the R lists I subscribe to when I saw a reference to the Mondrian visualization/graphics package. I was certain the author had made a mistake: Mondrian is a relational online analytical processing (ROLAP) engine, not graphics software. So I set out to search on Google and, lo and behold, found two Mondrians – the ROLAP package that OpenBI knows and loves, and the visualization platform mentioned in the email. I then downloaded and started fiddling with the free open source graphics -- and liked what I saw.
Though beta at the time of first review, Mondrian graphics is now in production, albeit version one. In addition to the software, product developers Martin Theus and Simon Urbanek recently published a companion book: Interactive Graphics for Data Analysis, Principles and Examples1. I was able to get an early copy and was impressed enough with its content and format to make both the text and software topics for a series in Dashboard Insight. The columns will present concepts from the book, including data analysis, in many ways a close kin to modern business intelligence (BI), and interactive graphics. They will also explore the Mondrian software and its place in a 2009 BI portfolio, especially one that’s budget-constrained.
Theus and Urbanek are computational statisticians, concerned more with computing than math. They dedicate their book to the late John W. Tukey, progenitor of Exploratory Data Analysis (EDA), now often simply called data analysis. Tukey’s work became prominent in the 70’s and 80’s, in some ways a reaction to the “mathematization of statistics” prevalent then. I’m a big fan of EDA and think Tukey was ahead of his time – and certainly far ahead of computing power available then to implement his ideas.
Whereas traditional or confirmatory statistical methods are concerned with testing hypotheses, exploratory data analysis obsesses on formulating hypotheses. EDA proponents note that “Exploratory Data Analysis lets the data speak to you without the interference of models” – simultaneously a pithy moniker and jibe at the statistical status quo.
No comments have been posted yet.