Eastern Analytical Symposium

E25: Chemometrics without Equations (or Hardly Any)

One-Day Course 
Date to be announced; 8:30am – 5:00pm
Dr. Neal Gallagher, Eigenvector Research, Wenatchee, WA

COURSE DESCRIPTION

Chemometrics without Equations (CWE) is designed for those who wish to explore the problem-solving power of machine learning tools without high level mathematics. Course emphasis is on proper development and interpretation of models as applied to real-life problems. The objective is to teach in a simple way so that participants will be better chemical data science practitioners and managers.

CWE, introduces Principal Components Analysis used for exploratory data analysis/pattern recognition, and regression methods Classical Least Squares and Partial Least Squares regression. Classification methods SIMCA and PLS Discriminant Analysis are also discussed. The course concludes with methods for model improvement with advanced preprocessing and variable selection.

WHO SHOULD ATTEND
Chemometrics without Equations was developed for chemists and bio-scientists who need to analyze multivariate chemical data and for those who manage these staff members. It covers the basics of pattern recognition, quantitative (regression) and qualitative (classification) models. Additional sections discuss how to improve models through advanced data preprocessing and variable selection. The course is relevant in a large number of analytical applications including process analysis, analytical chemistry, pharmaceuticals, sensory, medical devices, metabolomics, etc.

TOPICS

  1. Introduction
    1. What is chemometrics?
    2. Resources
  2. Pattern Recognition Motivation
    1. What is pattern recognition?
    2. Relevant measurements
  3. Principal Components Analysis
    1. What is PCA?
    2. Scores and loadings
    3. Interpretation
    4. Supervised and unsupervised pattern recognition
  4. Regression
    1. What is regression?
    2. Classical least squares (CLS)
    3. Inverse least squares (ILS)
    4. Principal components regression (PCR)
    5. Partial least squares regression (PLS)
  5. Classification
    1. What is classification?
    2. Classification based on PCA models: SIMCA
    3. Using regression for classification: PLS Discriminant Analysis (PLS-DA)
  6. Advanced Preprocessing
    1. What are the goals of preprocessing?
    2. Mean- and median-centering, autoscaling
    3. Normalization and standard normal variate
    4. Savitsky-Golay and filtering
    5. Generalized least squares weighting clutter suppression(GLS)
    6. Multiplicative scatter correction (MSC)
  7. Variable Selection
    1. Why do variable selection?
    2. Knowledge based selection
    3. Model based, e.g. on loadings
    4. Interval PLS (iPLS)
  8. Conclusions

ABOUT THE INSTRUCTOR
Dr. Neal B. Gallagher, is Vice President and co-founder of Eigenvector Research, Inc. established January, 1995. Recent research includes novel algorithms for iterative target detection in hyperspectral imaging and shift-invariant tri-linearity (SIT) for modeling hyphenated chromatography data. Neal has been intimately involved in chemometrics consulting, teaching short courses and software development including algorithms for detection, classification and quantification. Specific interests include hyperspectral imaging, process modeling, multi-variate curve resolution and classical least squares modeling.

Featured Posts