The Premier Chemometrics Company

November 2015
No. 8

Recent Happenings at Infometrix
Infometrix held a custom chemometrics training course this past September at GW Pharmaceuticals in the UK and a general three day course in October at our office in Bothell, WA with attendees from various industries. The course was well received and we look forward to our next general course offering in spring of 2016. Inquire and see how  you can benefit from this course. Dates will be announced soon.

In October, Infometrix attended the Gulf Coast Conference in Houston, Texas and presented a couple of papers. Abstracts are below and the full papers can be viewed on our website under the Publications page.

Abstract # 7 Paper, 30 minutes
"Re-engineering Calibration in Optical Spectroscopy"
Brian Rohrback, Infometrix, Inc.

Abstract # 173 Paper, 20 minutes
"Fast GC for True Control of a Process"
Brian Rohrback, Infometrix, Inc.

Thank you for your continuing subscription to our monthly newsletter. In this issue we begin a discussion about multivariate statistical process control and how Infometrix is involved in developing the science around product quality analytics and applying those methods to real time manufacturing processes. In future newsletters we will discuss real world applications that produce savings in time and reduce waste in the process stream.

Are you a Pirouette user? Below are some things you may not have known. We have had a demo version of Pirouette for many years. This isn't your typical demo. We discuss the advantages of a non-licensed copy of Pirouette and how it can benefit you and your work. Also, be sure to check out the tips column for discontinuous variable exclusions.

Multivariate Statistical Process Control

Statistical process monitoring and control is critical to all but the simplest manufacturing processes.  If you mix, cook, react, harvest, or otherwise torture the ingredient materials into a final product, you are better off monitoring process parameters in between the GoesIn and the ComesOut parts.  Next to support for specific analytical instrument systems and requests for specific multivariate algorithms, the most common request I hear is to incorporate the time dimension so that Pirouette models can be integrated into the process beyond just the end-point monitoring typical of the analytical instrument stage.

In partial answer to these requests, we did add a form of control charting to InStep, but it was designed mostly to provide a testbed environment for model makers to assess the effectiveness of MSPC based on a Pirouette model on-line.  It was handy, but no substitute for the robust systems available.  Instead of diluting our chemometrics effort, we sought a partner with SPC capabilities that wanted to add the M prefix to their product.  Our partner is Northwest Analytics (NWA –

So, as Lloyd Colegrove, Director of Fundamental Problem Solving at Dow Chemical, puts it: “In terms of dollar value, that depends on how you value the mistakes you avoid.”  Check out the video on Conquering Big Data in Chemical Manufacturing at the ARC Industry Forum 2013.

After Further Review...
In a quality control setting, if you deploy a reliable, automated chemometric process, there is tangible economic benefit to the organization.  If you do not bring this solution into routine use, playing with chemometrics is just a corporate expense.

Infometrix has been around for 37 years and in that time we have seen an amazing blossoming of progress in areas needed for chemometrics to be successful.  In particular, the interfaces between computers and the instruments they control have been very critical for the field.  Now that we are entering into the officially-announced BIG DATA realm, the issue of handling complex interactions of many instruments holds both huge promise and significant risk.  Are we up to the task?  Infometrix has participated in the commercialization of more than 100 products distributed by 30+ companies.  We have also created or advised in the creation and routine deployment of over 1,000 chemometric models of varying complexity.  The factors that lead to success fall into three categories: the instrument, the samples, and the deployment.  Half of the work lies in the model development; the literature adequately documents how the chemometric approach can be made to succeed in performing objective data interpretation of complex, multivariate data sets.  What is less-well documented, and where chemometrics systems most often fail, is in not making these applications easy to usher into routine use.  In addition, in order for a company to benefit fully from a chemometric method, the analytical experience gained in one location should be transferable in a plug-and-play manner worldwide.  The chemometric results must be immediately available whether a simple report, integration into the instrument software, fed to a SQL database, or messaged to your phone.

Over the course of the next few email newsletters, several applications will be detailed where the results of chemometric analysis is integrated into mission-critical QC processes.  Examples will be drawn from chromatography and spectroscopy and show how the processing can make use of a database or digital control system.

Brian G. Rohrback

Pirouette Demo - It's Much More
For nearly two decades, Pirouette has been available as demonstration software. Besides the obvious ability to examine before you purchase, the Pirouette Demo fulfills many purposes.
  • Defining chemometrics and an approach to data analysis. - First and foremost, we view the Demo as a means to explain chemometrics processes and learn the specific approaches we use in Pirouette to solve problems. The Demo does not come with a time limit and is supplied with full documentation, so those interested in learning more about the philosophy of chemometrics can go through the tutorials or examine the mathematics behind the algorithms. There are walk through exercises for both quantitative and qualitative analyses, and we insert many literature references that can further your understanding of a particular topic.
  • Free data viewer. - The Pirouette Demo also functions as a free viewer, so that an unlicensed user can see the results of someone else's Pirouette analysis by loading their PIR file and displaying all of the results they generated. It does not require a purchase to be able to review another person's work.
  • Raw data visualization and organization - A copy of the unlicensed Pirouette software will allow you to read and display many types of instrument vendor files and standard format files. Therefore you may use Pirouette's organizational and data visualization capabilities on raw data to look for unusual samples or find general relationships in your data.
  • Free teaching aid - The Pirouette installer includes a full copy of our help file, which is actually more of a text book with walk through exercises. Also included with the installation are a number of example data files, allowing you to fully process these "blessed" files, so you can gain hands-on experience with the Pirouette interface and chemometric processing. In this way, Pirouette can be deployed as a teaching aid. 
  • Translate data to standard formats - With the multitude of instrument file formats making data difficult to move from the dedicated instrument software into other programs, we opened up the ability to translate any data from one readable format to a supported output format (usually CDF, ASCII or Excel).
If there are ways a demonstration version of the software can serve your needs better, let us know. Of course, we are always interested in hearing your comments.

To leave comments or questions on any of the topics presented in this newsletter, please visit the Discuss page on our website and look under the title of this newsletter. Thank you for your support and continued readings.

In This Newsletter
-Pirouette Demo - It's Much More
-Upcoming Events
-Tech Tip: Excluding Discontiguous Variables within Pirouette
-After Further Review...

Upcoming Events
54th EAS
November 16-18, 2015

Peftec 2015
November 18-19, 2015

IFPAC 2016
January 24-27, 2016

Pittcon 2016
March 6-10, 2016

Tech Tip: Excluding Discontiguous Variables Within Pirouette
A unique feature introduced in Pirouette over 25 years ago is the ability to create subsets of a data file within the Pirouette session, and use the included data within the subset for additional data analyses. This gives the user an ability to perform ‘what-if’ interpretations on different segments of data in the file. These subsets can be made from the row space (subsets of samples) or the column space (subsets of variables) or both.
The process of creating subsets can be done from data tables (which can become unwieldy for large files) or from graphical views. Another way to prepare for subsets is to create class variables and to use the different values assigned to samples as a means to segregate groups. However, Pirouette does not have any easy way to mark variables as there is no equivalent to the class variable in the row space. But, with an extra step, it’s not too hard to do.
The procedure involves the use of a mask row in which you insert specific values to indicate which variables to keep. Conceptually, it helps to use a zero value for variables that you intend to exclude and a large value for those you intend to keep. For example, considering inserting a value an order of magnitude larger than any real value. You could even create the values in another program, such as Excel, then copy and paste them into a blank row in the Pirouette table.
After creating this dummy row, make a throw-away subset by selecting the row and choosing Edit > Create Include. Run PCA on this new data subset, then plot the Loadings. Those variables with large values will stand out obviously in the first factor direction. Highlight those (or everything else, depending on which set of variables you want to exclude) and do a Create Exclude. Name your new exclusion set and go from there.
When it comes time to do predictions, there is no further manipulation you need to make.  Pirouette models track the excluded variables automatically and will perform this trim prior to running the algorithm's prediction function.  You just supply prediction data in which all variables are included.

For more tips, visit the FAQs and Tips or User Questions page on our website.

Learning is like rowing upstream: not to advance is to drop back.

-Chinese proverb

The Infometrix mission is to provide high quality, easy-to-use software for the handling of multivariate data.

Publications - Application of chemometrics to problems in a variety of research areas

Eastern Analytical Symposium & Exposition 2015 -  Somerset, NJ
Nov 16-18, 2015

PEFTEC 2015 - Infometrix will present a talk at this  conference and exhibition specializing in monitoring and analytical technologies for the petroleum, refining and environmental industries.

IFPAC 2016 - Arlington, VA
Jan 24-27, 2016

Pittcon 2016 - world’s largest annual premier conference and exposition on laboratory science

Copyright © 2015 Infometrix, Inc., All rights reserved.