A big thank you to all the speakers and attendees who made the 2019 Analytics Solutions Conference a huge success! Below are the talks presented at the conference, with links to the pdf files of their slides.
Organizations face increasingly complex problems that are critical to their operation and, in some cases, for their survival. Such problems require the proper use of data and its interpretation. A major issue is how to develop appropriate solution strategies to develop good solutions efficiently and effectively.
Historically, data analysis focuses on tools: design of experiments (DOE), regression analysis, statistical process control, modeling. More recent tools form the foundations for analytics and data science. Tools are important for creating good solutions to complex problems. However, it is crucial to understand “the right tool, for the right job, at that right time, correctly applied.” Today, there are too many people who claim that their tool is the universal solution. The reality is more complex.
This talk outlines the new discipline that is devoted to the art and science for creating good solutions for complex problems using data. The paradigm for this new discipline is chemical engineering. Chemical engineering builds upon both chemistry and mechanical engineering to create new chemical processes and to improve existing chemical processes efficiently and effectively. Crucial to these solutions is the concept of “unit operations”. Chemical engineering theory focuses on understanding these core operations and how to develop proper strategies to deploy them. Our new discipline, statistical engineering, takes such an approach to the complex problems facing organizations today.
This talk introduces at a high level this new discipline. It then outlines the important roles that both DOE and analytics play in the solution of complex problems. In the process it emphasizes the importance of strategy and understanding exactly what the tools can and cannot do.
Partial least squares regression, which has been around for about four decades, is a dimension-reduction algorithm for fitting linear regression models without requiring that the sample size be larger than the number of predictors. It was developed primarily by the Chemometrics community where it is now ingrained as a core method, and it is apparently used across the applied sciences.
And yet it seems fair to conclude that PLS regression has not been embraced by some, even as a serviceable method that might be useful occasionally. Nor does there seem to be a common understanding as to why this rather enigmatic method should not be used, although bumptious discussions of PLS failings can be found in some applied areas. Perhaps this is as it should be — perhaps not.
This talk is intended as a relatively informal overview of PLS regression from a statistical perspective, including historical context, personal encounters, methodology, relationship to envelopes and, near the end, a few recent asymptotic results for high-dimensional regressions.
In a time where digital manufacturing transforms the flow of goods into a flow of data and raw materials, brand owners need effective ways to secure quality and protect their products against counterfeiting. New technologies increase speed and quality in distributed manufacturing, but add complexity to the supply chain … and possible security challenges. Intellectual property protection requires strategies to ensure that both the data and the goods are secure.
Non-destructive, speedy, and user-friendly field testing boosts both security and quality monitoring. New handheld instruments are cost-effective, especially when their capabilities are boosted with strong analytics. A blend of classic methods and new technologies using chemical tags, spectrometers and analytics provide protection for products threatened by counterfeiting, saving money and reputation.
This talk outlines the elements of the solution from the chemical “fingerprint” to field authentication using spectroscopy. This can be applied on pharmaceuticals, cosmetics, spare parts, electronics, wine and additive manufacturing, and the talk will include use cases and real-life examples. It will also touch on blockchain and when blockchain is helpful, and when it is just expensive hype.
"Know the SCOR for Multifactor Strategy of Experimentation": Mark Anderson (Stat-Ease)
"Founder's Favorite Features in Design-Expert Software - V12 and Beyond": Pat Whitcomb (Stat-Ease)
"Optimizing the Vapor Deposition of Bioactive Films using Sequential Experimentation": Lou Johnson (JMC Data Experts)
"Analyzing Experiments Involving an Amount Factor with a Zero Setting": Howard Rauch (Eastman Chemical)
"Practical Considerations in the Design of Experiments for Binary Data": Martin Bezener (Stat-Ease)
"Text Mining: Discovering Themes in Text Records": Paul Prew (Ecolab)
"Improving Collapsibility Robustness of an EPS-CD by means of Simulation and Six Sigma": Michal Majzel (ZF)
"Design of Experiments in Chemistry: the Pitfalls": James Cawse (Cawse and Effect, LLC)
"DOE: A Formulator's Perspective": William Arendt (Arendt Consulting)
"How to Increase Design of Experiments Success": Carol Parendo (Collins Aerospace)
"Using Experimental Design and Statistical Software to Investigate the Impact of Amines on Metalworking Fluid Lubricity": Jason Pandolfo (Quaker Chemical)
"Using Multi-Sensor Data Fusion for Process Analysis and Control": Geir Rune Flåten (Camo Analytics)
"PAT in pharmaceutical formulation manufacturing": Angela Spangenberg (DisperSol)
"The Importance of Flexibility of Multivariate PAT Techniques": Eric Jayjock (Patheon)
"PAT Best Practices": Chuck Miller (Camo Analytics)
"Introduction of Hyperspectral Image Analysis for Quality Control (new Camo Analytics product)": Geir Rune Flåten (Camo Analytics)
"DMicroNIR: PAT for Endpoint Assessment in Blending": Ed Gooding (Viavi Solutions)
"MVA and DOE: Throughout the Product Lifecycle": Chuck Miller (Camo Analytics)
"Multivariate Analysis: From Chemometrics Modeling to Process Analytics": Sylvie Roussel (Ondalys)