Statistical Analysis of Sensory Data collected from Sensory Panels using XLSTAT SENSORY and EyeOpenR
Sensory characterization of products has traditionally been collected from trained sensory panels, there is now an increasing use of “rapid methods” of data collection that do not require training. This course covers analysis of data from both sources. We assume only very basic statistical knowledge and there is a refresher session at the beginning to remind you of the basics.
The course uses XLSTAT SENSORY and EyeOpenR, as part of the course fee you will be given a complimentary licence to EyeOpenR software for 1 year.
Topics covered:
- Analysis of Variance review, standard errors, multiple comparison tests
- The Mixed model for sensory data, assumptions and problem data
- Panel Performance using ANOVA, the MAM model and simple multivariate methods
- Principal Component Analysis – how it works, interpretation of plots, covariance v correlation. Visualisation of sample differences, interpretation of maps
- Canonical Variates Analysis – a technique that better displays product differences v panel variation
- Rapid methods of data collection, Free Sorting Tasks and Napping, guidance in design and approaches to analysis.
- Relating sensory to instrumental/consumer liking 1: Simple regression modelling, variable selection. Modelling curvature, limitations.
- Relating sensory to instrumental/consumer liking 2: Partial Least Squares Regression, How it works, guidance in model building.
Course Outline
Module 1 – 23 February @ 07:30 GMT
Quick Stats Refresher: Precision of means, standard errors and confidence intervals, Least Significant Difference (LSD). Statistical significance, what it measures and what it does not measure.
Analysis of Variance of Sensory Panel Data using mixed models, multiple comparison tests. Assumptions and problem data. XLSTAT Product Characterisation.
Module 2 – 25 February @ 07:30 GMT
Principal Component Analysis(PCA). How it works, difference between covariance and correlation based analysis. Which numerical outputs are useful. Interpretation of graphics, supplementary variables and observations.
Canonical Variates Analysis (CVA) – for showing sample differences relative to panel variation
Module 3 – 2 March @ 07:30 GMT
Assessing Panel Performance using ANOVA attribute by attribute. Panel and panellist repeatability, discrimination and consistency. Setting action standards. XLSTAT Panel Analysis. Investigating scaling effects using the MAM model in EyeOpenR software.
Multiple Factor Analysis – a method for investigating links between data tables
Rapid methods Napping – Objectives of the method, design issues, analysis using XLSTAT
Module 4 – 4 March @ 07:30 GMT
Relating Sensory data to instrumental or liking data
Regression modelling refresher, multiple regression, variable selection, modelling curvature. Limitations
Partial Least Squares as a technique to build models to predict one block of data from another, e.g. sensory from instrumental data or liking from sensory. Graphical explanation of the technique and comparison with PCA. Key statistics and their interpretation, Strategies for model building.
Each Module of three hours is available to purchase separately as a stand alone session, or book on all four modules to get the discounted price and benefit from the full Hands on Sensory experience. When booking, please confirm which modules you wish to attend – either in the comment box of the booking form, or drop us an email at askqi@qistatistics.co.uk.