The methodology being developed in this project has many applications, for example:
1. process understanding and control
2. analysis and understanding of sensory data and consumer research (sensometry)
3. handling new challenges in biotechnology (bioinformatics).
User-friendly statistical methods wanted
Food products are always affected by a number of factors, such as raw materials and process conditions, for example. The properties of raw materials are affected by many very complex factors, such as those involving genetics or environment. These factors affect a whole range of the properties that determine a product's quality. To be able to understand such complex connections, and also to make use of this understanding in product development and control, it is vital to have statistical methods that can be used to analyse many complex variables and samples simultaneously.
Diverse applications
During the course of the project, Nofima Mat has developed several new and user-friendly methods. Including:
1. Validation techniques to control and counter the risk of over-optimistic results. The technique encompasses the whole mathematical process, not just estimating the statistical parameters but also such things as choice of model, selection of variables and linearisation.
2. Methods for correcting light scattering effects in single cells and in tissue that use background information from the light scattering theory in "soft modelling".
3. Methods for improving predictions in proteomics. A completely new method has been developed in which 2D gel images are analysed at pixel level, instead of the measurement of spot volume that is used in all commercial software. In this way we avoid the whole problem of missing values and we can even find variations for proteins that strongly overlap on a 2D gel and in the edges of saturated protein spots.
4. Techniques and software for studying inter-disciplinary data sets. This is one of the greatest challenges created by modern research and development, in which the really useful insights come from the connections between many types of data. We are working on developing improved techniques for the user-friendly presentation of results, something which is absolutely essential for effective communication between user and data analyst.
5. A strategy for analysing data with complex correlations that is suitable for use in situations with few samples and many variables, as is often the case in functional genomics. The object is to find interaction between genes.
6. PanelCheck, software for simple and very effective control of assessors in a sensory panel. The software was developed by Nofima Mat and is free to use. PanelCheck visualises the results of advanced statistical methods, making it easy for the user to interpret the results. This is done with the aid of graphs and plots and a simple and intuitive user interface.
7. Development of and actual applications for methods of combining sensory and other product properties in consumer surveys. Methods for connecting these two aspects are largely absent from present literature. We will develop methods and software for an area where methodology is largely absent and easily available software is totally absent.
8. Methods for weighting the assessors' contributions in a sensory panel, so as to achieve a more realistic presentation and description of the products tested. Weighting will be influenced by the assessors' performance, dependent on both performance parameters and how the individual assessor judges in relation to the rest of the panel.
