Data analysis is a vital part of science today and in assessing quality, modern data sets contain many variables where the most interesting trends are hidden within a combination of the variables. Using the latest multivariate techniques with design of experiments, participants will learn how to interpret complex data quickly and confidently. You will discover the secret of overviewing data tables and also learn how to build robust predictive models that translate data into decisions.
Multivariate analysis is used widely in many industries, from raw material analysis and drug discovery in the pharmaceutical industry, early event detection and gasoline blending in refineries, right through to predicting future market trends in business intelligence applications.
It can be used for measuring data sets with many input variables or for investigating the trends in time series data, all of which provide a better understanding of a given issue and often result in resource and time savings.
It can generate data models that can be used for on-line prediction and classification and it also generates data models for faster product and process optimization for application in Spectroscopy, Chemometrics, Sensometrics, Process Analytical Technology, Product Development and Quality Control.
For most real-world problems we need to use multivariate analysis to model the relationship to the response. We need tools to mine our data, and understand the relationships in them, whether for pattern recognition, or to develop models that can be used to predict values for new samples. Multivariate analysis (MVA) can be applied to different data including instrumental data, medical diagnostics, census data, economic data, marketing data, or even a sports team’s performance. MVA gives us a means to find the relationships in the data, and provides tools to visualize the relationships between samples and variables. It can be used for both qualitative and quantitative analysis.
With this 2-days hands-on training you will learn how to interpret complex data quickly and confidently using the latest software tools of chemometrics and multivariate analytical methods. Participants can quickly and safety start using the methods in their own work to optimize, classify or predict products and processes. Solutions 4U provides Camo’s Unscrambler software for the exercise session during the course.
Who should attend?
The course combines theoretical studies and practical workshops, ensuring that each participant gets individual focus, and understands the practical uses of MVA application. It is intended for researchers, scientists, chemists and engineers involved in process understanding, rapid quality monitoring, product formulation, metabolomics and sensory science or likely to work with spectroscopic instruments (NIR, FTIR, UV, UV/VIS, NMR, Raman, Mass Spectroscopy), chromatography instruments (LC, GC) and other sources of multivariate data as part of laboratory, R&D, quality control or production processes. No prior knowledge of statistics is assumed.
Note: Due to the nature of the course and the learning expectations, the availability seats are limited. You need to register early to obtain confirmation of your space.
No prior knowledge of The Unscrambler® is required to attend our courses.
- Introduction to MVA
- Multivariate Data
- Data Analysis & chemometrics in practice
- Multivariate Analysis workflow
- Uni-variate data, Statistics and Plotting
- Introduction to Multi-variate Analysis
- Data Collection and Design of Experiements
- Data Checking and Pre-processing
- Principal Component Analysis (PCA)
- Principles Theory of PCA
- Model rank and cross validation
- Score Plot, Loading Plot & Bi-Plots
- Detecting Outliers in PCA
- Importance of validation in MVA
- SIMCA Classification
- Multivariate regression:
- MLR- Multiple Linear Regression
- PCR – Principal Component Regression
- PLS – Partial Least Squares
- Interpreting a regression model in 5 steps
- Prediction Process
- Detecting and dealing with outliers
- Study impact in regression
- Hands-on exercise