Liking data analysis

Use this feature to analyze liking data quickly and efficiently. Available in Excel using the XLSTAT software.

Description of a liking data analysis

Liking data (also called hedonic data) are among the most collected in sensory analysis. They simply consist in asking the different subjects/consumers/assessors to give a score to the products, generally with a predefined scale on which they have to answer.

Even though the idea behind liking data is very simple, the analysis of these data is rich. The first step is a description of the liking data, with their distribution by product, the differences between sessions, and the visualization of the data... A second step, more advanced, consists in performing comparison tests between products and building an internal preference mapping. The last step is based on the study of the agreements between the assessors with the comparison or clustering of groups of assessors or the clustering of the latter.

Setting up a liking data analysis in XLSTAT

Liking data: Select the data corresponding to the different assessors. If the first row of the selection includes headers, the option "Variable labels" in vertical format or "Assessor labels" in the horizontal format must be activated. If you are in vertical format and you select several columns, they will be averaged.

If the format is horizontal:

Product labels: Check this option if you want to use the available product labels. If you do not check this option, labels will be created automatically. If a column header has been selected, check that the "Attribute labels" option has been activated.

If the format is vertical:

Products: Select the products corresponding to the liking data rows. If a column header has been selected, check that the "Variable labels" option has been activated.

Assessors: Select the assessors corresponding to the liking data rows. If a column header has been selected, check that the "Variable labels" option has been activated.

Sessions: Select the sessions corresponding to the liking data rows. If a column header has been selected, check that the "Variable labels" option has been activated.

Options of a liking data analysis in XLSTAT

Center the assessors: Activate this option to center the assessors (mean of each assessor set to 0).

Scale the assessors: Activate this option to scale the assessors (variance of each assessor set to 1).

Confidence interval (%): Enter the percentage range of the confidence interval to use for the various tests and for calculating the confidence intervals. Default value: 95.

Clustering of the assessors: Activate this option to cluster the assessors (see "Clustering of the assessors" section). Next, decide if you want XLSTAT to automatically define a truncation, and therefore the number of clusters to be retained, or if you want to define the number of clusters yourself.

Results of a liking data analysis in XLSTAT

Differences between sessions: The table of differences between sessions (standard deviations if you have more than two sessions) is displayed. It allows you to see the possible errors of the assessors or of the data entry (especially if a value is high).

Means of the differences between sessions for each product: The table of the means of the differences between sessions by product is displayed followed by the associated graph. These results can be used to determine if certain products have resulted in session differences.

Means of the differences between sessions for each assessor: The table of the means of the differences between sessions by assessor is displayed followed by the associated graph. These results can be used to determine if any assessors resulted session differences.

Data in horizontal format: Data in horizontal format are displayed without missing values (they are estimated by the option chosen). This data allows you to choose certain options yourself by entering them in an internal preference mapping, an Agglomerative Hierarchical Clustering...

If you have selected groups, the following results will be displayed group by group. In addition, if you have selected the "Center the assessors" option, some results will be given before and after centering the assessors.

Product means: The product means table and the associated bar graph are displayed. This result allows you to determine how much the products are appreciated.

Box plots of the liking scores by product : The box plots of the liking scores for each product are displayed. These allow you to visualize the dispersion of liking data within a product and to compare the dispersions between products.

Visualizing data : A graph allowing to visualize directly the data of the different assessors is displayed. You can choose the assessor to highlight in order to check its data or to compare it to others.

ANOVA: This table allows you to evaluate the explanatory power of the product factor. The explanatory power is evaluated by comparing the fit (as regards least squares) of the final model with the fit of the rudimentary model including only a constant equal to the mean of the dependent variable (liking data). In other words, if the p-value is significant, we reject the hypothesis that all product means are equal.

Means charts: These graphs allow you to visually compare the means of the products with the associated confidence intervals.

Product/Tukey (HSD): The results of multiple comparison tests of the product means are displayed to determine which products are different from each other and which are similar. The product groups are then given.

Internal preference mapping: The results of the internal preference mapping are displayed. They start with the eigenvalues of the factors as well as the percentages of inertia that each one represents, before displaying the coordinates of the assessors and the coordinates of the products. All these coordinates are also displayed in graphs. Note: if an assessor does not have a representation quality higher than 50% (sum of the squared cosines of the assessor on the axes > 0.5), then it is not displayed.

Differences for each product: The results of the ANOVA, the multiple comparison tests between classes, and the associated graphs are displayed for each product.

Clustering of the assessors: The results of the cluster analysis of the assessors are displayed. They contain the obtained dendrogram (truncated if the said option has been checked), and the assessor clusters built by truncating the dendrogram.