School of Food Science and Nutrition Faculty of Mathematics and Physical Sciences
Laboratory handbook 2019-20
FOOD5472M Sensory Science
Autumn and Spring terms
Module manager: Dr. Peter Ho
Teaching Staff:
Dr. Peter Ho (p.ho@leeds.ac.uk)
Contents
Contents 1
1 Introduction 2 1.1 What you are required to do before the practical session . . . . 2 1.2 Whatyoushoulddoduringapracticalsession . . . . . . . . . 2
2 Lab01: Introduction to Sensory measurement 3
2.1 Introduction……………………….. 3
2.2 MaterialsandMethods………………….. 3
Instructionsforteamspreparingsamples. . . . . . . . . . . . . 3 PartI.The3-AFCtest………………….. 4 PartII.TheSensoryranktest………………. 5 PartIII.ThesensoryratingtestwiththeLMS . . . . . . . . . 7
2.3 Dataanalysisandreportwriting …………….. 8
3 Lab02: Overall liking with the 9-point hedonic scale 10 3.1 Introduction……………………….. 10 3.2 Overview ………………………… 10 3.3 MaterialsandMethods………………….. 11 3.4 Dataanalysisandreportwriting …………….. 13
4 Lab03: Flavour profiling using Profile Attribute Analysis 14
4.1 Introduction……………………….. 14
4.2 Overview ………………………… 14
4.3 MaterialsandMethods………………….. 15
Referencesamples ……………………. 15 Step1.Trainingsession …………………. 15 Step2. Attributeidentificationsession . . . . . . . . . . . . . . 16 Step3. Individualassessmentofthesamples . . . . . . . . . . 16
4.4 DataanalysisandReportwriting…………….. 17
5 Stats01: Assessing multivariate sensory descriptive data 18 5.1 Introduction……………………….. 18 5.2 DescriptiveanalysisofFrenchwines . . . . . . . . . . . . . . . 18
1
CONTENTS 1 5.3 Dataanalysisandreportwriting …………….. 19
6 Stats02: Assessing the Performance of a sensory descrip-
tive analysis panel 21
6.1 Introduction……………………….. 21
6.2 Methods…………………………. 21
AnalysisofVariance …………………… 21
Exampleofalatticeplot…………………. 22
6.3 DataanalysisandReportwriting…………….. 23 PartI.Assessmentofindividualassessors . . . . . . . . . . . . 23 Part II. Assessment of whole panel and final selection . . . . . 24 Part III. Assessment of differences in the sensory profile . . . . 25
CONTENTS 1
1 Introduction
This handbook provides details of the sensory laboratory practical sessions that will be held during the autumn and Spring term.
1.1 What you are required to do before the practical session
• You should make sure that you have read through the instructions sheet found in this handbook.
• You should make sure that you have brought along a printed copy of the instructions or have access to this online document during the session.
1.2 What you should do during a practical session
1. Make sure that you have a laboratory coat with you when you enter the Food Technology lab
2. Try and arrive at least 5 minutes before the start of the session.
3. Register your presence by signing the yellow form
4. Demonstrators will be available to guide you during the session.
5. Follow the instructions from the demonstrator as to what you should do.
6. Make sure you have read the participation information sheet and signed the consent form before agreeing to taste any samples. You can decline to taste the samples, if you have a good reason to do so. e.g. Religious, presence of an allergen you might be allergic to. However, you should still take part in the preparation of samples. See the role of sensory analysts for each practical.
2
2 Lab01: Introduction to Sensory measurement
2.1 Introduction
The main aim of this practical class is to introduce you to three of the most commonly used sensory methods for collecting nominal, ordinal and interval data. The three sensory tests you will be examining will be
1. A 3-AFC test to determine if one sample is sweeter than another.
2. A sensory rank test for comparing differences between three or more samples
3. A sensory rating test using the LMS scale to compare the degree of sweetness between two samples or more samples.
2.2 Materials and Methods
The practical session will be divided into three parts, where each method will be examined.
Instructions for teams preparing samples
As an analyst team, you will be required to conduct all tests by
1. Your team will be responsible for preparing sufficient product samples for each of the tests for all the members of another team. You will be provided with a worksheet with the information regarding the sample presentation order, the three-digit codes for each sample and the assessor number associated with the coded samples.
2. Samples to be tested: You will be provided with samples of (to be pre- sented on the day of the test by the demonstrator). The demonstrator
3
CHAPTER 2. LAB01: INTRODUCTION TO SENSORY MEASUREMENT 4
will inform you which sample should be sample A, which one sample B and which one sample C.
3. Before an assessor is permitted to evaluate the samples, make sure that you have placed the required number of product samples, in the correct order for each separate tasting period, on the sample preparation side of each sensory booths.
4. Before the test is conducted, you will need to make sure that every assessor from the other team has read the information sheet, signed a consent form and has been properly instructed on the correct procedures for evaluating the samples.
5. Ask each assessor to carefully read the instruction on the computer screen before tasting any samples. See the sections below about the specific details that are required.
As an assessor, you will be required to conduct all tests by
1. preparing the samples for the test.
2. organise each sensory booth with the material required for this test. Each team will prepare the required number of samples for another team.
Part I. The 3-AFC test
In the 3-AFC test, the panellist is asked “Which is the sweetest out of the three samples presented? ” The three possible samples would be two B and one A (ABB, BAB, BBA).
Instructions for analysts
1. Each assessor should receive all three samples together.
2. Organise each sensory booth with the material required for this test, as described in the general instructions for preparing samples. Each team will prepare the required number of samples for another team.
3. Place all three sample and a cup of drinking water on the assessor’s side (serving counter) of the sensory booth. The test will be conducted using Compusense® five.
4. Ask each assessor to begin by carefully reading the instruction on the computer screen before tasting any samples.
5. Instruct each assessor to take a sip of water to cleanse their palate before tasting the first coded sample.
CHAPTER 2. LAB01: INTRODUCTION TO SENSORY MEASUREMENT 5
6. After that, instruct the assessor to taste each sample, from left to right, tasting each sample just once, followed immediately by filling in the on- screen questionnaire on the computer screen. Make sure you instruct the assessors not to take a sip of water between tasting each sample.
7. When they have finished allow them to cleanse their palate and they should be instructed to leave the room and will be instructed when they will be asked to return for the next session.
Instructions for assessors
You will be required to taste three samples. There will be only one tasting period.
1. You will be assigned an assessor number for the taste panel. Note down your number.
2. Before going in to the taste panel room, read the information sheet and sign the consent form.
3. Enter the taste panel room by the entrance for the assessor and take a seat on one of the chairs in a booth.
4. The sensory analyst on the other side of the booth will explain what you need to do.
5. Following the instructions on the screen. You must register for the ses- sion by selecting the panellist number in the list. Follow the instruction until you get to the screen with the first questionnaire.
6. Wait for the set of samples to be presented to you. Check that the sample code of the samples presented matches the code on the screen. If it is different inform the sensory analyst.
7. If it is OK. Follow the instructions on the screen.
Part II. The Sensory rank test
In a sensory rank test, each assessor will be presented with the full set of all three samples that are to be determined. For this second part:
Instructions for analysts
1. Each assessor should receive all three samples together.
2. Organise each sensory booth with the material required for this test, as described in the general instructions for preparing samples. Each team will prepare the required number of samples for another team.
CHAPTER 2. LAB01: INTRODUCTION TO SENSORY MEASUREMENT 6
3. Place all three sample and a cup of drinking water on the assessor’s side (serving counter) of the sensory booth. The test will be conducted using Compusense® five.
4. Ask each assessor to begin by carefully reading the instruction on the computer screen before tasting any samples.
5. Instruct each assessor to take a sip of water to cleanse their palate before tasting the first coded sample.
6. After that, instruct the assessor to taste each sample, from left to right, tasting each sample just once, followed immediately by ranking the sam- ples by filling in the on-screen questionnaire on the computer screen. Make sure you instruct the assessors not to take a sip of water between tasting each sample.
7. When they have finished allow them to cleanse their palate and they should be instructed to leave the room and will be instructed when they will be asked to return for the next session.
Instructions for assessors
You will be required to taste three samples. There will be only one tasting period.
1. You will be assigned an assessor number for the taste panel. Note down your number.
2. Before going in to the taste panel room, read the information sheet and sign the consent form.
3. Enter the taste panel room by the entrance for the assessor and take a seat on one of the chairs in a booth.
4. The sensory analyst on the other side of the booth will explain what you need to do.
5. Following the instructions on the screen. You must register for the ses- sion by selecting the panellist number in the list. Follow the instruction until you get to the screen with the first questionnaire.
6. Wait for the set of samples to be presented to you. Check that the sample code of the samples presented matches the code on the screen. If it is different inform the sensory analyst.
7. If it is OK. Follow the instructions on the screen.
CHAPTER 2. LAB01: INTRODUCTION TO SENSORY MEASUREMENT 7
Part III. The sensory rating test with the LMS
Each assessor will be presented with each sample separately and will be re- quired to rate the sample for the degree of sweetness using the labelled mag- nitude scale (LMS).
Instructions for analysts
1. Each assessor should receive only two samples.
2. Organise each sensory booth with the material required for this test, as described in the general instructions for preparing samples. Each team will prepare the required number of samples for another team.
3. Place the first sample and a cup of drinking water on the assessor’s side (serving counter) of the sensory booth. The test will be conducted using Compusense® five.
4. Instruct each assessor to take a sip of water to cleanse their palate before tasting the first coded sample.
5. Ask each assessor to begin by carefully reading the instruction on the computer screen before tasting any samples.
6. After that, instruct the assessor to taste the first sample and rate the degree of sweetness, using the LMS scale, by filling in the on-screen questionnaire on the computer screen.
7. When they have finished with the first sample, they should be told to click the next button on the screen and inform the sensory analyst that they have finished evaluating the first sample.
8. Remove the sample and replace it with the second sample.
9. Instruct each assessor to take a sip of water to cleanse their palate before tasting the second coded sample.
10. The assessor should then taste the second sample and follow the same procedures in rating as was used with the first sample. When they have finished with the second sample, they should click the next button on the screen and inform the sensory analyst that they have finished evaluating the second sample.
11. Remove the sample and replace it with the third sample. The assessor should then taste the third sample and again follow the same procedures in rating as was used with the first and second sample.
12. When they have finished allow them to cleanse their palate and they should be instructed to leave the room and will be instructed when they will be asked to return for the next session.
CHAPTER 2. LAB01: INTRODUCTION TO SENSORY MEASUREMENT 8
Instructions for assessors
You will be required to taste two samples. There will be two separate tasting periods.
1. You will be assigned an assessor number for the taste panel. Note down your number.
2. Before going in to the taste panel room, read the information sheet and sign the consent form.
3. Enter the taste panel room by the entrance for the assessor and take a seat on one of the chairs in a booth.
4. The sensory analyst on the other side of the booth will explain what you need to do.
5. Following the instructions on the screen. You must register for the ses- sion by selecting the panellist number in the list. Follow the instruction until you get to the screen with the first questionnaire.
6. Wait for the first sample to be presented to you. Check that the sample code of the sample presented matches the code on the screen. If it is different inform the sensory analyst.
7. If it is OK. Follow the instructions on the screen.
8. Repeat this procedure until you have tasted all samples.
2.3 Data analysis and report writing
The report should written using either the R markdown template file or the WORD .docx file. The use of the R markdown file is preferred, however, due to problems with ISS clusters not be able to run R markdown in Rstudio, either version of the template can be submitted. Follow the instructions in the report template for Lab01. The report should be submitted online no later than a week after the practical session, unless otherwise stated. Please refer to the online information for the correct deadlines. Please note that penalties for late submissions will be applied. Please refer to the Code of Practice on Assessment or you can log onto the Food Science and Nutrition minerva page under Organisations, where you can find this file and others.
The data collected for the sensory rating test with the LMS scale must be converted to the appropriate values before analysing the data. Rescaling the data by taking 10 to the power of the Compusense rating value. e.g, if the value from Compusense is 1.52, then 101.52 = 33.1, which is equal to the category “Strong” on the LMS scale. You can do this in Excel before during the analysis in Rstudio.
CHAPTER 2. LAB01: INTRODUCTION TO SENSORY MEASUREMENT 9
The marking criteria for the report can be found in the report template. Specific feedback will be provided on turnitin (part of Minverva) and general feedback, if required, will be provided during a tutorial or lecture session.
3
Lab02: Overall liking with the 9-point hedonic scale
3.1 Introduction
The practical session will introduce you to the use of the 9-point hedonic scale. Practical experimental procedures, including experimental design con- sideration and appropriate statistical analysis methods, are outlined. The procedures described can be used to determine the:
1. degree of overall liking of a single product or sample
2. degree of liking between two sample can be compared to determine if there is a preference for one or more over the other samples
3.2 Overview
Assessors are presented between two or more samples, where each sample is presented separately (one at a time), and are instructed to rate the degree of overall liking of each sample using a 9-point hedonic scale. Data is collected and the mean rating values for each test sample is used to determine if any one sample is preferred by the panel using an appropriate statistical test.
The paired t-test should be used when comparing just two samples to determine if there is a significant preference between the two samples. A single factor within-subjects analysis of variance (ANOVA) without an interaction term (also called a two-factor ANOVA) is used to determine if there are any significant differences between more than two samples.
Care must be taken to ensure that the ANOVA model is valid by (i) checking for the absence of a significant interaction effect between the sample and assessor factor: and (ii) checking for model violations, such as normality of residuals or non-consistency of error variance. Other violation, such as
10
CHAPTER 3. LAB02: OVERALL LIKING WITH THE 9-POINT HEDONIC SCALE 11
the presence of outliers should also be examined. If for example there is a departure from normality, then the raw data should be converted into ranks and the Friedman test should be performed. The Friedman test should also be used when the interaction term is significant or other model assumptions are violated.
If the null hypothesis from the ANOVA test or the Friedman test is rejected for the sample factor, then Tukey’s honestly significant difference (HSD) can be is used to determine any significant differences between the samples.
3.3 Materials and Methods
In this laboratory session, you will be required to
1. run the test by preparing samples and presenting them to assessors from a different group
2. taste sample that have been presented by another group to you, as an assessor.
You take turns being an assessor or sensory analyst (those presenting sam- ples).
Instructions for analysts
1. Yourteamwillberesponsibleforpreparingsufficientproductsamplesfor all the members of another team. You will be provided with a worksheet with the information regarding the sample presentation order, the three- digit codes for each sample and the panellist number associated with the coded samples.
2. Samples to be tested: You will be provided with samples of (to be presented on the day of the test by the demonstrator). The demonstrator will inform you which sample should be sample A, which one sample B and which one sample C. Note: Each assessor should receive only three samples (one sample from product A, one sample from product B and one sample from product C). Check the worksheet provided.
3. Before the assessors are permitted to evaluate the samples, make sure that you have placed the three different product samples, in the correct order, on the sample preparation side of each sensory booths. Now place the first sample and a cup of drinking water on the assessor’s side (serving counter) of the sensory booth. The test will be conducted using Compusenseà five.
CHAPTER 3. LAB02: OVERALL LIKING WITH THE 9-POINT HEDONIC SCALE 12
4. Before the test is conducted, you will need to make sure that every assessor from the other group has read the information sheet, signed a consent form and has been properly instructed on the correct procedures for evaluating the samples. The correct procedures for a test is to ask each assessor to begin to carefully read the instruction on the computer screen before tasting the samples.
5. They should be told that they will evaluate each sample separately. They should then take a sip of water to cleanse their palate before tasting the first coded sample. After that, instruct the assessor to taste the first sample and rate the degree of overall liking by filling in the on-screen questionnaire on the computer screen.
6. When they have finished with the first sample, they should click the next button on the screen and inform the sensory analyst that they have finished evaluating the first sample. Remove the sample. Replace them with the second sample. The assessor should take a sip of water to cleanse their palate before tasting the second coded sample.
7. The assessor should then taste the second sample and rate the degree of overall liking by filling in the on-screen questionnaire on the computer screen. When they have finished with the second sample, they should click the next button on the screen and inform the sensory analyst that they have finished evaluating the second sample.
8. Remove the sample. Replace them with the third sample. The assessor should then taste the third sample and rate the degree of overall liking by filling in the on-screen questionnaire on the computer screen.
9. When they have finished, allow them to cleanse their palate and thank them for their participation.
Instructions for assessors
1. You will be assigned an assessor number for the taste panel. Note down your number.
2. Before going in to the taste panel room, read the information sheet and sign the consent form.
3. Enter the taste panel room by the entrance for the assessor and take a seat on one of the chairs in a booth
4. The sensory analyst on the other side of the booth will explain what you need to do.
CHAPTER 3. LAB02: OVERALL LIKING WITH THE 9-POINT HEDONIC SCALE 13
5. Following the instructions on the screen. You have to register for the ses- sion by selecting the panellist number in the list. Follow the instruction until you get to the screen with the first questionnaire.
6. Wait for the first sample to be presented to you. Check that the sample code of the sample presented matches the code on the screen. If it is different inform the sensory analyst.
7. If it is OK. Follow the instructions on the screen.
8. Repeat this procedure until you have tasted all samples.
3.4 Data analysis and report writing
The report should written using either the R markdown template file or the WORD .docx file. The use of the R markdown file is preferred, however, due to problems with ISS clusters not be able to run R markdown in Rstudio, either version of the template can be submitted. Follow the instructions in the report template for Lab02. The report should be submitted online no later than a week after the practical session, unless otherwise stated. Please refer to the online information for the correct deadlines. Please note that penalties for late submissions will be applied. Please refer to the Code of Practice on Assessment or you can log onto the Food Science and Nutrition minerva page under Organisations, where you can find this file and others.
The marking criteria for the report can be found in the report template. Specific feedback will be provided on turnitin (part of Minverva) and general feedback, if required, will be provided during a tutorial or lecture session.
4
Lab03: Flavour profiling using Profile Attribute Analysis
4.1 Introduction
The Flavour Profile method was one of the first descriptive analysis meth- ods developed to measure a flavour profile of a food product. The method uses a small number of assessors that should be trained to evaluate reference samples that exhibit the sensory characteristics that are likely to be present in the products samples that are the focus of the sensory profile. This method requires that consensus be made between assessors in terms of what is per- ceived by the panel. To allow for the use of statistical analysis of the data, the flavour profile method has been modified so that a consensus is first made in terms of the type of attributes to be perceived by the panel, but ratings of at- tributes in the product samples are made individually. This method is known as Profile Attribute Analysis.These two methods are described briefly in the next two sections.
Practical experimental procedures, including experimental design consid- eration and appropriate statistical analysis methods, are outlined in the next section. The aim of this practical is to:
1. introduce you to the Profile Attribute Analysis methodology as a method to create a sensory profile of a product
2. represent the sensory profile, as a radar chart, which also indicates the analysis of variance (ANOVA) results for comparing any differences be- tween samples in the same radar chart.
4.2 Overview
The Profile Attribute Analysis has been modified from the Flavour Pro- file method to allow for the use of statistical analysis of the data. The panel
14
CHAPTER 4. LAB03: FLAVOUR PROFILING USING PROFILE ATTRIBUTE ANALYSIS 15
should work as a group to first identify and define the sensory attributes that they perceive. The aim here is to produce a general consensus amongst the group, with regard to the order in which the attributes should be rated. Asses- sors are required to first work individually in rating the samples. They should record any sensory attributes that they perceive, the order of perception, the Aftertaste and/or persistence, and finally assess the overall impression. Ref- erences samples can be examined to help the group in reach an agreement, regarding the final list of attributes to be used, their order of perception, the Aftertaste and/or persistence and overall impression.
After an agreement has been made as a group, assessors work individually to note the order of perception, the Aftertaste and/or Persistence, and finally assess the overall impression.
4.3 Materials and Methods
The experiment will be conducted in groups of 5 to 7 panellists. Divide your- selves accordingly and make sure that you have read the instructions on the training before starting. A demonstrator will be assigned as your panel leader
The procedure can be divided into three steps
1. A training session is firstly conducted to familiarise the panel with typ- ical flavours that might be found in the samples. Reference samples of individual flavours will be provided.
2. The second step is the attribute identification stage, where the panel discuss what they can perceive and the order of perception, using a reference sample similar to the samples in step 3.
3. In the third and final step, assessors evaluate the samples individually without the help from other assessors. They follow the procedures agreed upon from step 2 and provide their own ratings from the attribute list.
Reference samples
Each panel will receive a set of reference samples. Individuals should then take a small sample (about 30 ml) from each of the containers provided to the group. Make sure you label your containers so that there are no confusions as to which reference samples they are. You should also make sure that you have two empty containers, some napkins and some crackers or biscuits.
Step 1. Training session
Each panel member should taste each reference sample. You may taste the samples in any order, but make sure that you do not discuss what you perceived to any of your fellow assessors. The aim of this training session is
CHAPTER 4. LAB03: FLAVOUR PROFILING USING PROFILE ATTRIBUTE ANALYSIS 16
to help you recognise the individual flavours of each reference sample. Make notes of what you perceive as you taste the samples in table 1
You should take between 10 to 15 minutes to taste the reference samples. Repeated assessment of the samples will allow you to increase your ability to taste these samples
Step 2. Attribute identification session
When you have finished the training session, the panel leader will provide you with a small amount (approximately 20 ml) of each of the soup product samples. The aim of this session is to work as a group to identify the sensory attributes present in the soup. The soup would have been made from some of the reference samples in different amounts.
In this session, you should only identify the sensory characteristics that you can perceive. Write your answers down in table 2. Do not discuss the intensity of these attributes or which samples have a higher or lower amount of that attribute. You only need to note what can be perceived.
Discuss as a group and decide which attributes to use and their order of evaluation, as they are perceived. Each assessor should write down the full list of sensory descriptors, in table 3 provided in the worksheet, which will be used in the next step. You should take between 20 – 30 minutes to come up with your group’s list of attributes.
Step 3. Individual assessment of the samples
When the group has decided on the list of attributes that would be used for the flavour profile, each assessor should then rate each sample individually. Start with the first sample. As you taste the sample, note down the order of perception of each attribute. Start by identifying the first attribute by writing down the name of the sensory descriptor on the product sample rating sheet, provided in table 2 in the worksheet, and then use the 7-point rating scale, where 0 = Not present; 1 = just recognisable or threshold; 2 = weak; 3 = moderate; 4 = strong; 5 = very strong; 6 = Extremely strong, to rate the intensity of that attribute. Now continue with the identifying the next attribute and repeat the procedure of writing down the sensory descriptor and rating the intensity.
Follow the same procedure until you have rated all of the attributes, then use the same rating scale to rate the intensity of the after-taste, which is defined as a continuation of perceived flavours after you have swallowed the sample. You should then rate the persistence of the aftertaste.
Finally rate the overall impression on a three-point scale, where 1= low; 2= medium; 3 = high. When you have finished with the first sample, proceed to the next sample and follow the sample procedures as the first. Rate the final sample when you have finishing assessing the second sample.
CHAPTER 4. LAB03: FLAVOUR PROFILING USING PROFILE ATTRIBUTE ANALYSIS 17
4.4 Data analysis and Report writing
Collect the data from each of the assessors in your panel before you leave the laboratory. You should conduct your analysis in Rstudio and write your report, using the example report template from the tutorial, in Rmarkdown. Full details of R scripts (without showing the output in the pdf report) should appear in the appendix section at the end of the report. The main report should contain only the results of the analysis in the form of tables and figures and should not include any R scripts.
1. Conduct a single-factor within subject, at a 5% significance level, to determine if there are any differences between the samples for each at- tribute. Check the model assumptions for the parametric ANOVA and make any appropriate changes to required if the model assumptions are still invalid, use the or the Friedman two-way ANOVA and subsequent multiple comparison test. Decide, which results you should include in the radar chart.
2. Use the example script from the tutorial to create your own R script to create the correct data frame to plot a radar chart for each individual assessor.
3. Using the mean values for each sample for each attribute, create the radar chart for the whole panel. Put the results of the parametric or non-parametric ANOVA results in the plot (based on your conclusions from part [1.]. You should also create the appropriate table in your report
4. Discuss your results in the relevant sections of your report, similar to that shown in the example Rmarkdown document.
5
Stats01: Assessing multivariate sensory descriptive data
5.1 Introduction
Data collected from a sensory descriptive analysis panel on the assessment of products is multivariate. Simple radar charts can be used to compare products in terms of the different sensory attributes used to evaluate them. However, when the number of samples being compared is large, other methods are required. Additionally, the radar charts does not allow us to examine the relationship between variables. Multivariate analysis methods allow us to do what simpler methods are unable to do . A sample data set has been provided for the work required in this laboratory class. The aim of this class is to:
1. examine the relationship between variables using scatter plots and cor- relation tests
2. Use principal component analysis to analyse and present data
3. Conduct all analysis using Rscripts in Rstudio and write a report in Rmarkdown
The complete data set for this statistics laboratory, that can be found in the wine.RData workspace, will be required.
5.2 Descriptive analysis of French wines
The data comes from the FactoMinR package. Three different styles of of French wines from the Loire valley were analysed by a trained panel. The wines were from three regions (see the data column Label), Saumur, Bour- gueuil and Chinon. A total of 21 wines were assessed by the panel using 29 different aroma and taste attributes. The wines were grown in different soil
18
CHAPTER 5. STATS01: ASSESSING MULTIVARIATE SENSORY DESCRIPTIVE DATA 19
conditions (see the data column Soil), a Reference and Env1, Env2 and Env4. Samples were rated by each assessor on a line scale.
Table 5.1: Sensory descriptors used for assessment of wines
Characteristics Before Before Before Before Before Appearance Appearance Aroma Aroma Aroma Aroma Aroma Aroma Aroma Aroma Aroma Aroma Aroma Taste Taste Taste Taste Taste Taste Quality Quality Quality
Attribute Odor.Intensity.before.shaking Aroma.quality.before.shaking Fruity.before.shaking Flower.before.shaking Spice.before.shaking Visual.intensity Nuance Surface.feeling Odor.Intensity Quality.of.odour Fruity Flower Spice Plante Phenolic Aroma.intensity Aroma.persistency Aroma.quality Attack.intensity Acidity Astringency Alcohol Balance Smooth Intensity Harmony Overall.quality Typical
5.3 Data analysis and report writing
You should conduct your analysis in Rstudio and write your report, using the example R markdown report template from the tutorial. Full details of R scripts (without showing the output in the pdf report) should appear in the appendix section at the end of the report. The main report should contain only the results of the analysis in the form of tables and figures and should not include any R scripts.
1. Conduct an exploratory data analysis, using at least one of the three graphical methods that you have used in the tutorial to examine the re- lationship between each variable by comparing the relationship between the following descriptors:
1.1. Before 1.2. Aroma 1.3. Taste
Select the appropriate plot/s that allow you to explain the relationship between the variables. Discuss your results in your pdf document.
CHAPTER 5. STATS01: ASSESSING MULTIVARIATE SENSORY DESCRIPTIVE DATA 20
2. Conduct a pca analysis using all 29 variables and present a table of loadings and summary table.
3. Use one of two methods to plot a Screeplot to decide how many PCs to extract. Use the summary table to aid in your decision.
4. Plot the appropriate number of PC biplots for the following cases, show- ing differences based on
4.1. Region
4.2. Soil type
4.3. Region and Soil (Hint. Use the shape= option to use different symbols)
3. Discuss the results from the PC biplots, by examining the relationship between variables and relationship between variables. and the PCs
4. What differences were there in Aroma, Taste and Quality between the wines due to wine region and soil type?
5. Discuss your results in the relevant sections of your report, similar to that shown in the example R markdown document.
6
Stats02: Assessing the Performance of a sensory descriptive analysis panel
6.1 Introduction
The performance of a sensory descriptive analysis panel has to be monitored regularly to ensure that individual assessors and the whole panel are able to provide reliable and repeatable assessments of products. A sample data set has been provided for the work required in this laboratory class. The aim of this class is to:
1. examine data from individual assessors and the whole panel , using an Analysis of Variance (ANOVA)
2. using a lattice graph to aid in the interpretation of the ANOVA results
3. select only the appropriate asessors, based on their performance, to ex- amine the differences between samples for attributes used in a sensory descriptive analysis.
4. Conduct all analysis using Rscripts in Rstudio and write a report in Rmarkdown
The complete data set for this statistics laboratory, table4.xlsx, will be required.
6.2 Methods Analysis of Variance
Follow the instructions in the data analysis and report writing section on which analysis of variance models to use for each part of the analysis.
21
CHAPTER 6. STATS02: ASSESSING THE PERFORMANCE OF A SENSORY DESCRIPTIVE ANALYSIS PANEL 22
Example of a lattice plot
The ratings provided by each assessor for the rating of samples for each at- tribute can be examined by constructing a lattice plot. Figure 6.1 can be plotted using the Rscript.
library(ggplot2)
ggplot(table4, aes(Session,Springiness, colour= Sample)) +
geom_point(position=position_jitter(w=0.2,h=0.2)) +
facet_wrap(~Assessor, ncol=3) +
scale_y_continuous(breaks =seq(0,15, by=2))
This plot uses the ggplot2 library. The basic function used the data in the table4 R data object and plots Sessions on the x-axis and the ratings of Springiness on the y-axis. The parameter colour=Sample ensures that samples are represented with different colours. A “+” symbol is needed after that to then include any additional parameters that needs to be included. To plot points on the graph we use geom_point(). The parameters inside it allows use to make sure that any ratings that are the same are plotted not in the same spot but beside each other. This allows us to see both values. The next option included is facet_wrap(), which allows us to state that we want a lattice of different plots for each assessor. With this option, we can produce all of the required plots in a single plot. This is very useful, as we no longer need to construct each plot separately. Finally, I have included an option to indicate the number of tick marks for the y-axis, which corresponds to the 0-15 scale at intervals of 2 units.
Figure 6.1 shows us the differences between the rating of Springiness. As- sessor 9 is consistent between sessions, as all samples are rated in the same order and their is very little differences in the rating scores between sessions for the same sample. This is the same for Assessor 5. However, Assessor 5 indicates that sample 1 has the lowest Springiness, whereas Assessor 9 rates sample 1 as the highest and significantly higher as well. This is a clear example of a significant Assessor: Sample interaction effect. However, if we look at the assessment made by the other assessors, most of them seem to find that sam- ple 1 has the highest value of Springiness. Therefore, there the problem seems to be with the ratings from Assessor 5. What could be the reason for those differences. We should examine the ratings made by Assessor 5 for the other attributes to see if there are any similar observations in terms of differences in ratings between assessors. We also see an example of a Sample:Session effect for Assessor 1. Here we see that samples are not consistently rated at the same level between sessions. Other assessors seem to show similar behaviour but the effect might be minor and may or may not affect the ANOVA results for the whole oabel. A comparison of ANOVA for each assesor could help us explain what we see in this plot.
CHAPTER 6. STATS02: ASSESSING THE PERFORMANCE OF A SENSORY DESCRIPTIVE ANALYSIS PANEL 23
1
2
3
12 10 8 6 4 2
12 10 8 6 4 2
12 10 8 6 4 2
Sa
4
5
6
mple 1
2
3
7
8
9
123123123
Session
Figure 6.1: Springiness of three low-fat Cheddar cheese samples. Overlap- ping points have been jittered to aid viewing
6.3 Data analysis and Report writing
You should conduct your analysis in Rstudio and write your report, using the example R markdown report template from the tutorial. Full details of R scripts (without showing the output in the pdf report) should appear in the appendix section at the end of the report. The main report should contain only the results of the analysis in the form of tables and figures and should not include any R scripts.
Part I. Assessment of individual assessors
This part examines the ability of each individual assessor. Use the example Rmarkdown template to set up the data analysis section. You are required to do the suggested analysis for each individual assessor, however, you only need to provide the R script in the appendix for the analysis of one assessor.
1. Examine the performance of each individual assessor by conducting a two-factor (Sample, Session) ANOVA and the Tukey test for non- additivity for all of the sensory attributes.
2. Determine (i) individual discriminatory ability where p-value <0.5 for the FSample factor is acceptable; (ii) individual scoring reproducibility
Springiness
CHAPTER 6. STATS02: ASSESSING THE PERFORMANCE OF A SENSORY DESCRIPTIVE ANALYSIS PANEL 24
where a p-value <0.05 for FSession is unacceptable; (iii) individual scor- ing consistency where a p-value <0.05 for FSample x FSession interaction is unacceptable. Note that checking for model assumptions can be done to help explain some of the results, but no observations should be re- moved and refitting the ANOVA model to get a valid model is not neces- sary. This is because accessors performing well, based on the evaluation of the three criteria should not result in an ANOVA model that does not fit the model assumptions.
3. Construct a single table which compares the performance of each indi- vidual assessor, using the results from the analysis conducted. The table should include the relevant F values and indicate those values that do not meeting the criteria (see lecture for example).
4. Construct a lattice plot to compare the consistency of ratings for each assessor between sessions for each attribute.
5. Which assessors do not meet all of criteria for each attribute? Discuss your results in the relevant sections of your report, similar to that shown in the example R markdown document.
Part II. Assessment of whole panel and final selection
This part examines the results of the whole panel. Use the example Rmark- down template to set up the data analysis section.
1. Examine the performance of the whole panel by conducting a three- factor (Samples, Sessions, Assessor) ANOVA for all of the sensory at- tributes.
2. Determine (i) panel discriminatory ability where p-value <0.05 for the FSample factor is acceptable; (ii) panel scoring reproducibility where a p-value <0.05 for FSession is unacceptable; (iii) agreement between asses- sors where a p-value <0.05 for FAssessor. Note that checking for model assumptions can be done to help explain some of the results, but no observations should be removed and refitting the ANOVA model to get a valid model is not necessary. This is because accessors performing well, based on the evaluation of the three criteria should not result in an ANOVA model that does not fit the model assumptions.
3. Construct a single table which compares the performance of the whole panel, using the results from the analysis conducted. The table should include the relevant F values and indicate those values that do not meet- ing the criteria (see lecture for example).
CHAPTER 6. STATS02: ASSESSING THE PERFORMANCE OF A SENSORY DESCRIPTIVE ANALYSIS PANEL 25
4. Examine the table for the performance of individual assessors and re- move any assessors that do have the required individual discriminatory ability for the rating of more than two attributes. Keep those that are able to discriminate samples for all attributes and those that might have been unable to discriminate samples for not more than one attribute. Remove any further assessors that have did not have individual scoring reproducibility or individual scoring consistency for more than one at- tribute. After you have removed those assessors, repeat steps [1.] to [3.]
5. Compare the ANOVA results of the original panel with the ANOVA re- sults after removing poor performing assessors. Discuss the differences in the relevant sections of your report, similar to that shown in the ex- ample R markdown document. The panel without the poor performing assessors would be used for part III.
Part III. Assessment of differences in the sensory profile
The part of the study will be conducted on the panel with the assessors that have been selected on their performance only
1. For each attribute, conduct a single factor within-subjects ANOVA with replication to determine if the mean value of at least one sample is significantly different from another at a 5% significance level.
2. Check the model assumptions and make any appropriate measure to ensure that you have a valid ANOVA model. You might be required to remove sample replicates from the assessors. However, if you find that you have an assessor that has more than 2 replicates from the same sample that is responsible for model assumption violations, you should remove all the data from that assessor for the particular attribute. Re- analyse the data and if the model is valid, conduct a multiple comparison test at a 5% significance level.
3. Construct a single summary table that shows the mean values for each sample for every attribute that fitted the ANOVA model and their group- ings based on the multiple comparison results.
5. Which assessors do not meet all of criteria for each attribute? Discuss your results in the relevant sections of your report, similar to that shown in the example R markdown document.