This week we discuss the importance of choosing the right analytics problem, with a guest blog from Elder Research, Inc., a data science and analytics consulting and training company, with whom we have just joined forces. Our course spotlight is on:
- Feb 14 – Mar 13: Design of Experiments
See you in class!
– Peter Bruce, Founder and President
Choosing the Right Analytics Problem
The “streetlight effect”:
The streetlight effect. In June 1942, a version of the streetlight effect appeared in a popular syndicated comic strip, Mutt & Jeff. Three of the five panels in the strip are shown. The gentleman in the top hat is Jeff, and in the final omitted panel the officer joins Jeff in the search beneath the streetlight
This is related to the more general “Statistical Type 4 Error” – asking […]
Mining Your Own Business: A Primer for Executives on Understanding and Employing Data Mining and Predictive Analytics by Gerhard Pilcher and Jeff Deal
This is a short book, befitting its intended audience – managers and executives with responsibility for data science and analytics projects. It outlines the requirements for success […]
Statistics.com + Elder Research, Inc
As noted elsewhere, Statistics.com has now joined forces with Elder Research, Inc. (ERI). I have known Dr. John Elder, IV for well over a decade. He and I share an interest in statistical resampling methods, and John is a fellow author in data analytics, or, as it was called back when we became interested in the field, “data mining,” with books on data mining, text mining and ensemble methods. Elder Research’s consulting work is highly respected, combining technical expertise and managerial know-how to assure success with intellectual integrity. They won’t sell you shiny toys that are not part of a solid technical and business plan. It is an honor to be a part of the ERI team.
In the era of big data, some data remain small because they are very costly and time-consuming to obtain. Computer chip manufacturing, for example, is a complex chemical and photo lithographic process that can take months. Because manufacturers are always pushing the envelope on capability and size, reject rates from a batch tend to be high and there is a premium on tweaking the process to improve outcomes. There are many variables, and ad-hoc modification of parameters often increases variability, rather than improving the process. A rigorous experiment is required to know what really works and what doesn’t, but testing parameters two at a time in a simple A-B test would be far too time-consuming and costly. Instead we can vary multiple factors at once, using various design blueprints that have evolved in the field of industrial designed experiments. Learn more in:
Your instructor is Dr. James Rutledge, whose many years of consulting on designed experiments equip him to answer virtually any question. In the words of Robert Wells, Director of Business Intelligence at Adventist Healthcare:
“This class was very interesting and practical. I learned how to structure experiments to be able to isolate the effects that the input variables have on the response variable.”
You will learn how to
- Produce plots from your results
- Deal with interactions
- Specify fractional and full factorial designs
- Specify specialized designs, e.g. Taguchi, Box-Wilson, others
- Use Excel-based software to design experiments and analyze data
See you in class!
Digital badges provide employers and peers concrete evidence of what you have learned and the skills required to earn your credential. Each badge’s digital image holds verified metadata describing your qualifications and the mastery required to earn them.
Contact Us To Register or Learn More
If you have any questions on our courses, certificates, and degree programs and how they can apply to you, your work, and to your career, please get in touch. We’re here to help you succeed.