Pages: 8

Sensory and consumer science aims at measuring and understanding consumers sensory perceptions as well as the judgment, emotions, and behaviors that may arise from these perceptions. As in many scientific fields, the development of sophisticated statistical techniques and access to powerful data analysis tools have played an important role in the evolution of sensory and consumer science. In saturated and ever changing markets, however, being able to understand individual needs and preferences is critical should one intend to develop customized products. The way consumer and sensory data is explored today is extremely advanced and went way beyond the simple statistical analysis performed a few years ago using the data collected from standard consumers or trained panel studies. Rapid data acquisition, allied with the need for flexible, customized, and fast result interpretation, is opening a huge way for automation.

Pages: 24

R was created by Ross Ihaka and Robert Gentleman at the University of Auckland, New Zealand, and is currently developed by the R Development Core Team. There are several reasons why the people should learn R or any programming language for that matter. If most additions/improvements have a fairly limited reach, the introduction of the tidyverse in 2016 by H. Wickham revolutionized the way of scripting in R for many users. In R, variables can be called in different ways when programming. Version control is a tool that tracks changes to files, especially source code files. Version control systems are simply software that helps users manage changes to source code over time. In R, variables can be called in different ways when programming. Version control is a tool that tracks changes to files, especially source code files. Version control systems are simply software that helps users manage changes to source code over time.

Pages: 10

In this chapter, the authors explain what is data science and discuss why data science is valuable to sensory and consumer scientists. A more recent development in the history of data science has been the realization that the standard outputs of data science such as tables, charts, reports, dashboards, and even statistical models – can be viewed as tools that must be used in the real world in order to be valuable. One of the most important ideas in data science is that of reproducible research. From the standpoint of classical statistics, experiments are conducted to test specific hypotheses and proper experimental design ensures that the data collected will allow hypotheses of interest to be tested. It is worth noting that these two movements were connected by substantial work in the areas of statistical computing, knowledge discovery, and data mining, with important work contributed by Gregory Piatetsky-Shapiro, Usama Fayyad, and Padhraic Smyth among many others.

Pages: 30

In sensory science, different data collection tools may provide the same data in different ways. Also, different statistical analyses may require having the data structured differently. Another simple transformation consists in reorganizing the data set, either by reordering, adding, and/or removing columns. In some cases, new variables need to be created from existing ones. Examples of such situations include taking the quadratic term of a sensory attribute to test for curvature or simply considering a new variables as the sum or the subtraction between two (or more) others. It can happen that some columns of a data set contain information (strings) that cover different types of information. In some cases, the new column to create depend directly on the value(s) of one or more columns present in the data. Reshaping the data itself is done through pivoting which allows transitioning from a long and thin table to a short and wide table and vice versa.

Pages: 39

One of the main goal of data analysis is to produce results. Tables and graphs are the two fundamental vehicles to communicate information clearly and effectively. The various pre-attentive attributes that can be used to draw audience’s attention quickly and create a visual hierarchy of information include attributes of form, such as line length, line width, orientation, shape, size, added marks, and enclosure, attributes of color, which would be hue and intensity, and also spatial position and motion. When multiple plots should be generated using the same pattern on subset of data, it is possible to generate them automatically using facet wrap or facet grid. There are various ways to save or export ggplot charts. Currently, the background is gray with vertical and horizontal white lines, the legend is positioned on the right side, the axis is defined based on the data itself, there is no title, etc. Similarly to the theme, aesthetics can also be adjusted.

Pages: 25

Automating the analysis process is simple by applying the same code to different data sets. Effective communication of results is among the essential duties of sensory scientists and so is data collection, data preparation, data analysis, etc. Most data collection software offer the possibility to export data and results in Excel, while most data analysis software accept Excel format as inputs. Throughout the years, PowerPoint became one of the main supports for presenting results, whether it is in academia, in conference, or in companies. The process for building Word document directly from R is very similar to the one for PowerPoint, since it is also handled though officer. However, if the report is very standard or running routinely, R could automatically run the analysis as soon as new data is available, build the report, and send it automatically to the people, their manager, or colleagues and partners by email.

Pages: 5

The study was part of the project BISENS funded by the French National Research Agency. These biscuits were developed for breakfast consumption and specifically designed to improve satiety. The presentation order of the different products was randomized across the panel. Again, consumers evaluated one biscuit type per day/session. The time spent in the booth and the number of biscuits eaten by each participant were recorded by the experimenters, as well as the type of drink they selected and the approximate volume they drank during each session. In summary, the consumer test data have 1070 observations (10 Products × 107 Consumers) and 32 variables (including Consumer and Product identifiers). The biscuits consumer test.xlsx file also includes one sheet with consumption time and the number of biscuits that consumers ate, and a sheet indicating the average weight for each product.

Pages: 26

This chapter approaches all the required steps to reach such goal, from setting up the test (e.g. estimation of the number of panelists, design of sensory evaluation sessions and design of experiments), to the collection of data (through valuable execution tips) and its importation in a statistical software. Sensory and consumer science relies on experiments during which subjects usually evaluate several samples one after the other. Sensory and consumer science relies on experiments during which subjects usually evaluate several samples one after the other. For any sensory experiment that implies the evaluation of more than one sample, first-order and/or carry-over effects should be expected. Sensory and consumer scientists may sometimes consider using incomplete designs, that is, experiments in which each judge evaluates only a subset of the complete product set. Full factorial designs are of course commonly used and their application is usually straightforward.

Pages: 28

The goal of this chapter is hence to perform some pre-check of the data and to prepare them for future analyses. Data Preparation includes all the steps required to ensure that the data match their intrinsic nature. For instance, the attribute Light has one detected missing value. Rather than removing missing data, another strategy consists of imputing missing values. However, since most imputation methods involve modeling, applying them to variables with a high missing values rate can introduce bias in the data. Nominal data are any data that are defined through text or strings. To illustrate the conversion from character to numeric, let’s start with creating a tibble with two variables, one containing strings made of numbers and another one containing strings made of text. It is worth noticing that the individual differences could also be included by simply adding the Judge effect in the model.

Pages: 40

In fact, some data transformation can be seen as a part of both data transformation and data analysis. This chapter is somewhat more dedicated to the analysis of data by presenting how some of the most common analyses in sensory and consumer science are performed, integrating the analysis part to a script, and most importantly providing applications and extensions to all the procedures. The evaluated biscuits appear to differ the most for Initial hardness, Shiny, Dairy flavor, External color intensity, and Thickness. The biscuitstraits.xlsx file contains descriptive information regarding the consumers and their food-related behavioral traits. The analysis of consumer data usually involves the same type of analysis as the ones for sensory data, but the way the data are being collected and their underlying nature require some adjustments. There are various solutions for clustering depending on the type of distance, the linkage, and of course the algorithm itself.

Pages: 16

This chapter approaches important topics to help the people reach the goal of successful communication, from understanding the distinction between different analysis and audience awareness to methods to communicate, the art of storytelling, and ultimately, reformulate. Sensory and consumer scientists often act as consultant whether it be for their own company or for customers. One of the most important things about being a successful consultant is Audience Awareness. The general audience usually refers to the ones that are likely the final users or are somehow related, contributed, or are interested in the project. Something important to keep in mind is the follow-up process after a report is sent or a presentation is delivered. The ability to receive feedback and reformulate is undoubtedly a very important and sometimes challenging skill that consultants seeking success should be aware of.

Pages: 23

This chapter introduces and demystifies the concept of ML, by: explaining the differences between supervised and unsupervised ML models, proving that the people were already doing it long ago, perhaps without knowing, extending it to more advanced techniques, highlighting its main applications in the field. The notion of ML is vast, as it covers a large variety of analyses. In fact, ML algorithms are often classified based on the goals of their analysis. Three main groups are often considered: unsupervised learning, supervised learning and semi-supervised learning. In sensory and consumer science, unsupervised learning models are mainly used for Dimensionality Reduction and for Clustering. There are many clustering approaches and algorithms. They can be categorized into different types including exclusive, hierarchical, and probabilistic. Regression methods approximate the target variable with a combination of predictor variables. There are several ways to look into the model quality, with the approach and model metrics highly dependent on the situation/type of model used.

Pages: 22

Humans exchange information through the use of languages. There is of course a very large number of different languages, each of them having their own specificity. In recent years, open-ended comments have gained interest as it is the fastest, safest, and most unbiased way to collect spontaneous data from participants. Open-ended comments, and more generally textual responses in questionnaires, are by definition qualitative. Languages are complex, as many aspects can influence the meaning of a message. For an even deeper cleaning, let’s go one step further and split the remaining tokens into single words by using the space as separator. Check-All-That-Apply can be seen as a simplified version of open-comments in the sense that respondents also associate products to words; however, they lose the freedom of using their own as they need to select them from a predefined list.

Pages: 12

For simplification purposes, the code developed for this application requires that the data set contains one column called Judge, one column called Product, and all the other columns being quantitative. The user interface (UI) is the part of the application that controls what the user sees and controls. The server side of the application is where all the computations are being performed, including the construction of tables, figures, etc. Fortunately, many additional packages that create HTML tables have been developed to provide alternative solution to build more flexible tables. However, R provides other options that creates interactive graphs directly. Ultimately, shiny can also be combined to other tools such as rmarkdown to build interactive tutorials, teaching material, etc. The server side of the application is where all the computations are being performed, including the construction of tables, figures, etc. Fortunately, many additional packages that create HTML tables have been developed to provide alternative solution to build more flexible tables.

Pages: 3

More generally, the emergence of Generative Artificial Intelligence (AI) cannot and should not be ignored, as it can be a great digital assistant in many situations. However, be aware that the AI-generated code might be far from perfect, and expertise and understanding of coding language is extremely valuable to ensure that what is done is correct. Be also aware of any other of its limitations including privacy before using it. DiSTATIS is used to analyze multiple distance matrices collected on the same set of observation. It tackles the characterization of the products, panel performance assessment, links between sensory and instrumental data, consumer’s preferences, napping evaluation, and optimal designs. The analysis include ANOVA, posthoc tests, Clustering, Correlation, multivariate analysis, among many others.