There is always the hope that a new book*, even a product manual, will provide insights into those long forgotten (or never really grasped) fields that crop up again and again in one’s everyday life and work. Geoscientists and engineers of all persuasions are constantly confronted by data and expected to extract meaning from it. At the simplest level this is easy—we regularly cross plot this against that to extrapolate, interpolate and what have you. But there often remains a nagging suspicion that we have left something out of our analysis. Is the correlation significant? Is there enough data from which to draw a conclusion at all? The book ‘Analyzing and Interpreting Continuous Data’ (A&ICD) is a step by step guide to the use of SAS’ JMP statistical package to do just that—to seek out causality in data sets and back up one’s conclusions with a demonstration of their significance.
According to Wikipedia, JMP stands for ‘John’s Macintosh Project’ after JMP’s original author John Small, a rather embarrassing factoid that A&ICD glosses over. The toolset has application in the oil and gas vertical. JMP’s design of experiment functionality was used by Shell to design its cyclic steam stimulation program on the Peace River Carmon Creek tar sands project. SAS uses the tool as a front end to its ‘PAM’—Preventative Asset Maintenance (Oil ITJ November 2008).
A&ICD provides examples from a wide range of fields including manufacturing. The most immediate oil and gas application is in the downstream, as witnessed by the glowing endorsement from Dow Chemicals’ lead data miner and modeler Tim Rey. But anyone who is looking at a cross plot of anything—from seismic velocities to net to gross ratios is an implicit user of these techniques.
The book fills the promise of educating the reader—in part because it is derived from a US National Institute for Science and Technology (NIST) online resource** covering much of the same material—but without the SAS/JMP focus.
Not being specialists, we can’t say how original the material in A&ICD. But it has a ring of authority about it. We started with a dip into the Overview of Exploratory Data Analysis (EDA) which sets the scene with a vivid discussion of the philosophy behind the science behind the statistics. This is definitely not padding, but a serious look at what we are really trying to achieve—before setting out to explain how to achieve it. The same pattern repeats in each chapter, with a look at the problem description, key questions and tools—then an in depth discussion of the topic in hand. Only after this scene setting is the JMP toolset use discussed.
A&ICD starts with a introduction to basic statistical concepts, then works through EDA as above before moving on to characterizing manufactured products and materials.
There is a distinct bias towards discreet manufacturing—there is no mention of design of experiment in A&ICD, let alone of geostatistics. But if you use, or have wondered about using fancy statistics, be they six sigma, variance analysis or whatever, in any context, then this book will provide a thorough background in the underlying math and science that you really ought to have before leveraging these techniques and software tools.
* Analyzing and Interpreting Continuous Data Using JMP, A Step by Step Guide. José and Brenda Ramirez. SAS Press 2009. ISBN 978-1-59994-488-3 and www.oilit.com/links/1011_21
** The NIST handbook is available online at www.oilit.com/links/1011_20
© Oil IT Journal - all rights reserved.