Statistics for Sensory Analysis
Product code: SEN1-3
Duration: 2 - 3 days
What you will gain from this course
This course focuses on descriptive sensory profile data. You will learn how to run and interpret the main statistical techniques used to analyse sensory profiling data. You will understand the purpose of each technique and how various techniques can be used in combination. We also consider some of the methods for linking and modelling average consumer acceptability to sensory data with the aim of understanding drivers of liking and predicting likening of future prototypes.
Only the most basic statistical knowledge is assumed.
We offer the training in three one day modules, any one of these can be run on its own or combined with the other modules into a two or three day training course.
As part of the training package we supply a free copy of SENPAQ©, our own software for analysing sensory profile data, for each delegate. The training also requires a general statistical software package. We can advise on appropriate software packages.
We can customise the module content to meet specific requirements.
Analysis of Sensory Panel Data (one scale variable at a time)
- Precision of a mean – standard error and confidence interval
- Panel noise variation – interaction v panellist repeatability
- Comparison of pairwise means – t test
- Assessment of variability in means – F test
Analysis of Variance
- How it works
- Comparison tests and LSD’s
- Assumptions – which of my sensory variables will not give valid test results?
- How to deal with these problem attributes
Methods of assessing panel performance
- Three key measures: repeatability, discrimination and consistency
- Setting targets for panel performance
- Identifying problem panellists
Analysis of Sensory Panel Data (using many scale variables together)
Introduction to multivariate data
- Multivariate Data displays
Principal Component Analysis (PCA)
- How many underlying sensory dimensions are there in my profile?
- Producing product maps using PCA
- Interpreting the map
Canonical Discriminant Analysis
- Visualising product differences relative to panel variation
- Grouping products using sensory similarities
Generalised Procrustes Analysis
- Overlaying data matrices to form a consensus map
- Applications to panel performance and free choice profiling
Linking Sensory and Consumer Liking Data, Consumer Sensory Measures
Simple Regression Modelling
- Predicting liking from key sensory variables, identifying key drivers
- Principal Component Regression
Partial Least Squares Regression
- How it works, applications
Sorting and Napping Tasks
- Analysing data from free sorting tasks
- Napping methods and analysis using Multiple Factor Analysis2 –