- All Sections
- Article: Applications Library (1)
- Article: Articles (1)
- Article: Blogs (2)
- Article: Books (1)
- Article: Columns (41)
- Article: News (4)
- Article: Product (5)
This is Tony’s last column for Spectroscopy Europe. It is explores an idea that he has been developing for over 30 years, although as Tony points out the story starts around 3500 years ago.
Tony (A.M.C.) Davies looks at Multiple Linear Regression (MLR) this issue, as well as expressing his opinions about Principal Components Regression (PCR) and Partial Least-Squares (PLS).
Tony (A.M.C.) continues down the last furlong of his series of Tony Davies Column articles. This issue, he considers principal component analysis (PCA). Using research recently published in the Journal of Near Infrared Spectroscopy, he explains that PCA is a very useful tool but it will not solve all our problems. Two old articles on PCA, including Tony’s concept of the “Data Cake” have been added to the website, are referenced in Tony’s article and can be freely downloaded by readers.
Tony (A.M.C.) Davies starts a review of the chemometric ideas that have most excited him over the last 30 years. In this column, he looks at the use of Fourier transformation for data compression. FT can also provide the side benefit of reducing high-frequency noise.
Tony (A.M.C.) Davies and Tom Fearn describe an enhancement of the popular partial least squares (PLS) technique, powered partial least squares (PPLS), that has shown significantly better results.
Tony Davies makes sure we understand “What IS and what is NOT chemometrics” and why it matters.
Examining diffuse reflection and transmission spectra more thoroughly: Part 2. Sample temperature and sample composition
The Tony Davies Column once again contains a contribution from Karl Norris, who is widely regarded as the “father of NIR spectroscopy”. Karl continues to produce innovative ideas about the field and this article is no different. Building on the previous article concerning fourth derivatives, Karl has investigated the effect of varying gap sizes with some remarkable results.
Camo Software has announced the launch of a new version of its “all-in-one” multivariate data analysis and design of experiments software, the Unscrambler X. The new version provides improved performance and is optimised for better usability, security and reliability. Version 10.2 continues the company’s tradition of delivering multivariate data analysis software that is easy to use and offers good data visualisation. New features include the addition of Support Vector Machine Regression, Compliance Mode has been added as an installation option for those needing to meet the requirements of electronic signature and record handling, Recalculate with New has been added as a new feature to PCA, PCR and PLS, which allows the addition of new data to an existing model, Data Signature has been added for enhanced data integrity, Blocking of Designs has been added to Full Factorial designs in the Design of Experiments module, three new spectroscopy vendor specific imports have been added—DeltaVu, rap-ID and Visiotec—and an Interpolation function has also been added.
This Tony Davies Column is contributed by Karl Norris, known to many as “The father of near infrared spectroscopy”. He introduces his calculation method for fourth derivatives and shows how it can be used to extract instrument noise.
Tony (A.M.C.) Davies is again telling us to “Always look at the spectrum”. This time he uses an example from the development of a NIR spectrometer to demonstrate that one doesn’t really know what’s going on until one has “looked at the spectrum”.
Tony (A.M.C.) Davies stresses the importance of always looking at the spectrum, even if you [think you] know there’s nothing to learn. He relates his experience with noise in NIR spectra and what he has learnt from it. He would like us all to examine spectra for abnormality before relying on automated methods.
Umetrics announces the release of version 12 of FabStat, which combines multivariate techniques with semiconductor FDC know-how based on the SIMCA-P+ multivariate platform. It is a stand-alone application designed to seamlessly integrate into existing fab data management systems. It monitors the entire production line—each tool, wafer by wafer. When a fault occurs, it instantly identifies the out-of-control tool and points to the culprit signals. Features of this new version include handling of hierarchical models, supporting PLS models, supporting PLS-DA models and improved log messages.
Tony Davies and Tom Fearn present “A digression on regression”. They turn their attention to one of the simpler regression techniques, Classical Least Squares (CLS). As well as an explanation of the basics, they explain why it is not often used in spectroscopy, and give the pros and cons of various regression techniques.
A team of chemists from the University of Seville, Spain, has managed to distinguish between different kinds of tea leaves on the basis of their mineral content and by using artificial neural networks. This technique makes it possible to differentiate between the five main varieties of tea—white, green black, Oolong and red tea.
I recently “discovered” a very interesting radio programme on BBC Radio 4. It is “devoted to the powerful, sometimes beautiful, often abused but ever ubiquitous world of numbers”. A few weeks ago we were asked to say what we were doing while listening to the programme. The next week we were told that nearly 2000 e-mails had been received and this data had been given to information designer David McCandless to turn into a graphic. When this was trailed I got the impression that something new and exciting was going to be displayed and I thought that the graphic would include sound. The graphic is good but rather “ordinary” and I was disappointed. This got me thinking about how we display information. Have we made any advance in the last 25 years? Could sound be used!
Analyze IQ Ltd has released Version 1.2 of the Analyze IQ chemometrics software suite, which works with Windows 7 (32- and 64-bit) and earlier versions of Windows. A number of new file formats are supported, including JCAMP-DX, SpectroML and AIQ. New features include a model driven paradigm that separates model-building from model usage, so that you can package and deploy chemometric knowledge and analysis and storage of mixtures data.
Comprehensive chemometrics: chemical and biochemical data analysis
Editors in Chief: Steven D Brown, Romà Tauler and Beata Walczak
Elsevier 2009, four volumes, 2896 ppISBN: 9780444527028€1360
Reviewed by Chris Burgess
Spectroscopists, like chromatographers, produce large quantities of raw data. Data reduction into something more meaningful is the goal. Selecting the right tool for the job is the challenge. Inevitably the right tool involves heavy duty mathematical and statistical expertise and hence chemometrics.
Version 4 of the Pirouette software from Infometrix has been enhanced to support process monitoring and control applications. An update of the installer was also required to handle Windows 7 32-bit and 64-bit operating systems.
Something has happened to my data: potential problems with standard normal variate and multiplicative scatter correction spectral pre-treatments
A.M.C. DaviesNorwich Near Infrared Consultancy, 75 Intwood Road, Cringleford, Norwich NR4 6AA, UK. E-mail: firstname.lastname@example.orgTom FearnDepartment of Statistical Science, University College London, Gower Street, London WC1E 6BT, UK. E-mail: email@example.com
It is common practice in near infrared (NIR) calibration to apply pre-treatments designed to correct for the scatter effects usually seen in absorbance data. These pre-treatments can interfere with interpretation of the spectra. This is illustrated here with the aid of two rather extreme artificial examples.