Ph.D.-course
The Analysis of Sensory and
Consumer Data
at
Technical
University
of
Denmark,
Lyngby,
Denmark
September
9
–
September
13,
2013
Organized
by:
DTU Compute
Technical University
of Denmark
Course Objectives
To improve
the ability of analysing human perception data. Some of the newest
statistical methodologies will be covered using the open source
software R, among others the packages sensR, ordinal and lmerTest will
be used together with the PanelCheck software.
Learning Objectives
A
student
who has met the objectives of the course will be able to:
- Work with R and Panelcheck
- Plan and analyze simple discrimination and similarity experiments
using sensR
- Analyze replicated discrimination data
- Perform and understand simple Thurstonian modeling
- Use mixed models for sensory profile data and consumer preference
data by PanelCheck and R-packages.
- Analyze sensory profile data with the new scale correction method
(using R)
- Use PanelCheck for simple analysis as well as multivariate
analysis (including Tucker-1)
- Analyze A-not-A and Same-different data
- Analyze ordinal human perception data using the R-package ordinal
Language
All lectures will be given in
English.
Organizers
Per Bruun
Brockhoff, B324, R220, (+45) 2044
1711,
perbb@dtu.dk
Rune H.B. Christensen,
B324, R220, rhbc@dtu.dk
Programme,
overview
Monday: Simple
discimination using R (package sensR)
Tuesday: Multivariate
Analysis
using
PanelCheck
Wednesday: Mixed
models
using
PanelCheck
AND
R. (newly developed
R-routines)
Thursday: Advanced discrimination using R (packages sensR and ordinal)
Friday: More on replicated (discrimination/similarity) data analysis.
Brief student presentations.
In
addition an introduction to R is given and discussions/supervision
of participants individual data projects are included.
We begin Monday morning at 9am and finish Friday at 3pm.
Programme,
in more detail
Day 1, Monday:
Simple discimination using R (package sensR)
We
present the R-package sensR as a tool for the planning and analysis of
sensory discrimination and similarity experiments. The sensR package
includes easily accessible tools for handling the five basic sensory
test protocols: duo-trio, triangle, 2-AFC, 3-AFC, tetrad test. For all
of these sensR provides:
- Hypothesis tests
- Confidence intervals
(Standard and improved - likelihood based)
- Power and sample size
calculations
- Simulation
- Thurstonian analysis
- Plotting features
- Exercises based on given or own
data.
Day
2,
Tuesday: Multivariate Analysis using
PanelCheck
- Installation and PanelCheck GUI
(Graphical User Interface)
- Structure of QDA data
- Data import
- Performance indices (new in
PanelCheck)
- Plots based on one-way ANOVA /
one-factor fixed model
- Plots based on Principal
Component Analysis (PCA)
- Simple plots and methods
- PanelCheck workflow
- Export of plots in presentations
(PowerPoint)Excercises
- Analysis of a given or own data
set (groups of 2 to 3 students).
- Preparing a presentation with
results and conclusions from PanelCheck analysis
- Present results and conclusions
to other students (5 to 10 minutes)
Day
3,
Wednesday: Mixed models using PanelCheck
AND R. (Package lmerTest and other R-routines)
-
From
oneway
and
twoway
ANOVA
to 3-way mixed model ANOVA (PanelCheck)
- Correcting for scale effects in
sensory profile data (specialized R-routines)
- Using
the newly developed R-package lmerTest for the (automated) analysis of
more complex structured sensory and consumer data such as:
- Unbalanced sensory profile data
(e.g. missings)
- Incomplete consumer preference
data
- 2- (or higher)way product
structure in sensory
- 2- (or higher)way product
structure in consumer (Conjoint)-
- Extending Conjoint to include
Consumer background/design factors/covariates
- Complex blocking, product
replication, product batch structures in as well sensory as consumer
- A mixed model approach for
performing external preference mapping
- Extending
mixed model external preference mapping to include product and consumer
background/design factors/covariates (segments)
- Exercises based on given or own
data.
-
Day 4, Thursday: Advanced discrimination using R
(packages sensR and ordinal)
- Similarity tests for 2-AFC,
3-AFC, Duo-Trio, Triangle and Tetrad data
- Power and sample size estimation
for similarity tests
- Advanced test protocols: A-not
A, Same-Different, 2-AC
- Ordinal based discrimination
protocols: A-not A with sureness and Degree-of-difference test.
- ROC curve estimation and AUC
- Exercises based on given or own
data
Day 5, Friday: More on replicated (discrimination/similarity) data
analysis.
- Beta-Binomial
(standard
and
corrected)
analysis
for replicated data
- Replicated
Thurstonian Model for discrimination analysis
- Simulation of
replicated difference tests
- Replicated
categorical ratings/ordinal data more generally
- Exercises based on given or own
data.
- Brief student presentations.
Teachers:
Professor Per Bruun Brockhoff
Post doc Rune H.B.
Christensen
PhD Student Alexandra Kuznetsova
DTU Compute, Technical
University of Denmark
Research Scientist, PhD Oliver
Tomic, Nofima Mat, Ås, Norway
Participants
The course is designed for Ph.D.
students within
Statistics/Data analysis with interest in human perception data
and
Ph.D. students within non-statistical areas such as sensory
science,
food science, marketing etc. with interest in data
analysis and
statistics.
Study Material, by Teaching Days:
(It would be a
good idea to orientate yourself in the given material, although parts
of it indeed most likely is a bit technical for the typical participant
of this course - then focus on the example parts.)
Day 1:
Chapter 7 and 11 of: T. Næs, P.B. Brockhoff and O. Tomic, (2010).
Statistics for Sensory and Consumer Science, John Wiley & Sons.
Brockhoff, P.B. and Christensen, R.H.B. (2010). Thurstonian models for
sensory discrimination tests as generalized linear models. FQP, 21(3),
330-338.
Day 2:
Chapter 3, 4, 5 and 14 of: T. Næs, P.B. Brockhoff and O. Tomic, (2010).
Statistics for Sensory and Consumer Science, John Wiley & Sons.
O. Tomic, C. Forde, C. Delahunty, T. Næs, Performance indices in
descriptive sensory analysis - a complimentary screening tool for
assessor and panel performance, Food Quality and Preference 28 (2013),
122-133
Day 3:
Chapter 5, 8, 12 and 13 of: T. Næs, P.B. Brockhoff and O. Tomic,
(2010). Statistics for Sensory and Consumer Science, John Wiley &
Sons.
Kuznetsova, A., Christensen, R.H.B., Bavay C. and Brockhoff, P.B.
(2013). Automated Mixed ANOVA
Modelling of sensory and consumer data. To be submitted to: FQP.
Brockhoff, P. B., Schlich, P., & Skovgaard, I. M. (2012).
Accounting for scaling differences in sensory profile
data: improved mixed model analysis of variance. Submitted to: Food
Quality and Preference.
Day 4:
Chapter 7 of: T. Næs, P.B. Brockhoff and O. Tomic, (2010). Statistics
for Sensory and Consumer Science, John Wiley & Sons.
Brockhoff, P.B. and Christensen, R.H.B. (2010). Thurstonian models for
sensory discrimination tests as generalized linear models. FQP, 21(3),
330-338.
Christensen, R.H.B. Cleaver, G. and Brockhoff, P.B. (2011). Statistical
and Thurstonian models for the A-not A protocol with and without
sureness, FQP 22(6), 542-54.
Christensen, R. H. B., H.-S. Lee and P. B. Brockhoff (2012). Estimation
of the Thurstonian model for the 2-
AC protocol. Food Quality and Preference. 24(1), 119-128.
Day 5:
Chapter 7 of: T. Næs, P.B. Brockhoff and O. Tomic, (2010). Statistics
for Sensory and Consumer Science, John Wiley & Sons.
Christensen, R. H. B. and P. B. Brockhoff (2013). Analysis of
replicated categorical ratings data from sensory experiments. To appear
in Journ. of SFdS.
Christensen, R. H. B., H.-S. Lee and P. B. Brockhoff (2012). Estimation
of the Thurstonian model for the 2-
AC protocol. Food Quality and Preference. 24(1), 119-128.
Evaluation and Diplomas
To pass the full 2.5 ECTS
course, active
participation in all
activities is required INCLUDING the submission and
approval of a report subsequent to the course period. Grades:
Pass/Fail. ECTS points: 2.5
Registration
Email Camilla Lund Poulsen, DTU
Compute, Technical University of
Denmark, Building 324, DK-2800 Kgs.Lyngby, Denmark.
E-mail: capo@dtu.dk
Registration fee
For PhD (and
potential master) students: No fee. For others: 400 Euro for the
whole course and 100 Euro for single
day participation. (A discount for Sensometrics Society members will be
given)
Housing
Accommodation in hostels/hotels
is to be
arranged (and covered) by the participants themselves.
See the Visit
Copenhagen website at http://www.visitcopenhagen.dk/.
Updated pre-course information
VENUE and course start:
Monday 9. September, Room S12, Building 101, DTU Lyngby
All the teaching will take place in this room, which is a part of
the DTU meeting center in the main university building - called
building 101:
http://www.dtu.dk/english/About/Practical-information/Directions/DTU-Lyngby-Campus
We begin teaching activities at 9am, but invite you to register and
enjoy some breakfast between 8am and 9am.
R software
As is clear that apart from Tuesday (using PanelCheck Software) we will
use R in the course. We will introduce R as a start Monday, but we
recommend that you (try to) install the newest version of R AND Rstudio
on your laptop (so please bring one - we do NOT supply computers for
participants and they ARE indeed necessary) before you come -
please see the following short intro (with relevant links) to get
started:
http://02402.imm.dtu.dk/enote/afsnit/NUID145/
If you want to watch a few very short videos on how to get started with
R then you can go to this Youtube Playlist:
http://www.youtube.com/playlist?list=PLOU2XLYxmsIK9qQfztXeybpHvru-TrqAP
(A number of short videos illustrating the use of R - shown within a
Mac version,
but apart from the appearance of the sub-windows exactly the same works
for RStudio
in any operating system)
Individual Project Part of
SummerSchool:
To complete ("pass") the 2.5ECTS course 02930 you must at the end of
(or shortly after)
the course submit a brief report or presentation of some data analysis
using methods/tools from the course on your own data.
So please prepare to bring some data for the course. On the Friday
afternoon we have scheduled 90 minutes of participants (brief)
presentations of this. This individual (OR group based, if relevant)
project activity is planned to take
place in parallel with the organized program, in the following way:
- Monday/Tuesday: All participants will have a
one-to-one meeting with one of teachers of the course: Per Bruun
Brockhoff, Rune H.B. Christensen, Alexandra Kuznetsova or Christine
Linander. In this meeting the participant will present the data (and
background) to the "expert" with the aim to identify a suitable
mini-project for the course.
- Wednesday/Thursday: Participants work a little on
their mini-project in parallel with other activity (under supervision
by course teachers)
- Friday: 5-8 minutes presentations in plenum by
participants.
- After course: submit brief report or presentation
Basic Statistics Brush-up??
Finally, if you think that you need to brush up a little on parts of
your basic stats before coming,
you may want to have a look at some of the around 100 online modulized
basic stats videos by Per Bruun Brockhoff:
http://02402.imm.dtu.dk/podcast/english-lecture-recordings-e12/
Or find the same video-collection in iTunes-U by searching for "DTU
Statistics"