Courses

Since 2004 Corpus specializes in the development of

diagnostic- and analytical data analysis software solutions.

Some of Corpus projects also include data-aqcuisition. Common techniques are near infrared (NIR) spectroscopy and imaging.

Course 1: What to do when you lose overview of your data? Get back in control by multivariate transformation and visualization!

General

Many scientific studies and industrial processes are capable of producing huge amounts of data and one quickly loses overview. The course introduces methods to obtain insight in large to mega amounts of data and to regain control by easy visualization

1.Data files and formats

The starting point of all data analysis is numbers available in files. A good starting point is to learn how different file formats present the data and to go from one format to another one.

2.Getting a quick overview of the raw data by using different types of plots.

An easy way of getting a first overview of a large data sets is to visualize the whole set or parts of it. Different techniques can be used. A visualized data set is a map; reading and understanding this map is the only way use the data in a meaningful way.

3.Preprocessing of data

All data gains from preprocessing for easier interpretation and some basic techniques for preprocessing are fun-damental knowledge obtained here.

4.Visualizing statistical parameters from the raw data.

Once the raw data are mapped and visualized, statistical parameters can be visualized to create visual summaries of the data.

5.What is structure and what is noise?

All data contain different types of noise originating from different sources. Therefore, it is important to separate the noise form meaningful and useful structures. Noise can be visually described and interpreted.

6.Multivariate models for data exploration.

When data are correlated, a simplification can be made by calculating a multivariate model. A number of these models are introduced and their visual interpretation is explained and trained.

Who should attend?

Scientists, students and industry professionals who are interested in improving their existing data analysis skills. Participants can bring their own data for practicing.

Course format

One-day course consisting of theory lectures, demonstrations, practice using software.

Obtained result

The course participants will learn to handle data sets, data files, mapping of the data and a minimum of multivariate analysis.

Course 2: A systematic study of multivariate space by visualizing its structural components.

General

All large data sets are only structures in multivariate space and there is an easy pathway from raw data (a file) to a multidimensional visualization of this space giving a separation in meaningful structures and noise structures. Tables of statistical properties of the data and figures showing relationships inside the data can be made availa-ble for interpretations and decision making.

1.Visualization of the raw data to get an overview

Date file formats and how to extract an overview map. How to use this map to quickly get beforehand infor-mation on the data.

2.Data preprocessing

All data needs some form of preprocessing. Some preprocessing methods are statistically based and some are based on the physics of the data generating equipment.

3.Multivariate data analysis

The preprocessed data form a structure in multivariate space and there are methods (algorithms) to simplify this structure.

4.Interpretation of multivariate results as tables and figures.

Once a multivariate space model is made, it can be interpreted by looking at tables and figures. Making the cor-rect figures and tables and interpreting them is of the utmost importance.

5.Studying and removing noise components

Data analysis may become an iterative process where noise components are removed after a first model is made and then a new model is constructed from the modified data set.

Who should attend?

Scientists, students and industry professionals who are interested in improving their existing data analysis skills.

Course format

One-day course consisting of theory lectures, demonstrations, practice using software

Obtained result

The course participants will get a deeper understanding of multivariate space and how to interpret the models made on it. Participants can bring their own data.

Course 3: How to get good data: The only two ways possible ex-perimental design and correct sampling.

General

Through an effective way of planning and carrying out experiments, which is called "The Statistical Experimental Design Principle", the experimenter is given an opportunity of achieving:

•Optimal results

•A better choice of parameter settings

•A better understanding of the process studied

•Statistical significance of results obtained

A definition of robustness

In some situations, with large amounts of heterogeneous materials no experimental design is possible and cor-rect sampling is needed.

1.Introduction to Design of Experiments and the duality with sampling

A basic understanding of why and how of experi-mental design is needed. It is equally important to know how not to design experiments. Screening designs are easiest to make and understand (all calculations can be done by hand if needed) but they are powerful. An introduction to some relevant model diagnostics is a must. Factorial designs, frac-tional factorial designs and Plackett-Burman designs are included. All necessary nomenclature is includ-ed.

2.Screening Designs and Demo.

The importance and usefulness of some screening designs is demonstrated

3.Diagnostics

Some tabulated and visual diagnostics are ex-plained.

4.Response Surface (RSM), including demo diagnostics

In order to obtain detailed response surfaces, other than screening designs are required. The complexity of the models increases, but the information in the results goes up too. The most popular response surface designs are introduced and explained: Multi-level designs such as the composite, Box-Behnken and hexagonal (Doehlert) are explained in detail.

5.Mixture designs

For the special case of chemical mixtures with a fixed sum of concentrations the designs become triangular/tetrahedron and design space is con-strained. Also the calculation of the results and the diagnostics become special.

6.D-optimality

D-optimal designs are made when design space is constrained or when extra runs are needed.

7.Sampling in the laboratory and industry.

Who should attend?

Scientists, students and industry professionals who are interested in avoiding basic errors in experimentation and sampling by applying a more systematic approach.

Course format

One-day course consisting of theory lectures, demonstrations, practice using software

Obtained result

The course participants will learn how to avoid sampling traps and how to construct experiments systematically in order to obtain certifiable results.

Copyright Corpus Data & Image Analysis AB 2018 @ All Rights Reserved