Recurrence Quantification Analysis in the R Environment 2500-EN-F-202
The study of human behavior has often been based on essentially static
measures of performance (e.g. averaged reaction times, task accuracy)
and group comparison over these averaged measures. But human
behavior is very complex and dynamical, happens in time and within a
rich social network, and those static measures many times might just
oversimplify this reality. For this reason, we need to find new concepts
capturing the complexity of human behavior, and new analytical methods
which can deal with the dynamical nature of data collected, e.g. during
ecologically valid interactions.
While many of the theoretical concepts needed are to be found in
the theory of complex systems, which has been presented also in other
courses offered at WISP and which we will only shortly recap, this course
will concentrate on the way dynamical (i.e. time dependent) patterns of
behavior are inter-related and how we can extract key dynamical
measures from them. To this end we will offer an in-depth treatment of
Recurrence Quantification Analysis (RQA), a method for studying
dynamical (potentially nonlinear) time series data.
We encounter examples of this kind of data in many different
experimental or empirical settings in the cognitive sciences: from EEG,
eye-tracking and other physiological signals to more abstract and coarsely
categorized classes of behavior like in observational quasi-experiments
within a psychotherapeutic session; from linguistic streams during a
conversation to postural movements, and many more. In all those
situations the measured behavioral states are ordered temporally and are
rich in details – and not just simply indexed by a single performance
measure – and hence the multidimensional, continuously changing nature
of human behavior is fully represented.
RQA retains all the richness conveyed by the data and allows to
determine important dynamical qualities of the behavior, like its
stationarity, stability, the coupling with other signals (like in social
interaction), the complexity, the moments of transition etc. RQA has
found application in the study of language, deception, coordination, eye
movements, infant-mother interactions, and many of these studies will be
presented, commented and taken as a starting point for class discussions.
The course will not only discuss the principles and theoretical
background of the method but will mainly teach it in a practical way. We’ll
see how to clean the data first, how to set the parameters, run the
analysis and finally how to interpret and visualize the results. Students
will work on short home assignments and much of the work in class will
be on clarifying doubts and solving the encountered problems. Students
already working on their master thesis or on specific research projects are
also welcomed to try on their data the possibilities given by RQA.
Given the applied, hands-on approach advocated, a basic but
reliable knowledge of R is needed and will be the focus of the first few
classes. While R is a generalist statistical computing environment (it is the
most used among data scientists) we will give a practical but necessarily
limited idea of its open-ended possibilities, focusing soon on the functions
we will need the most in our treatment of RQA. Yet we hope that this
short introduction will encourage students in knowing more and using R
for other related statistical tasks.
Type of course
Learning outcomes
Learning outcomes
Students will be able to perform many basic but important operations
over data within the R statistical environment: importing the data,
understanding the basic data types, visualizing and summarizing the
data.
Students will understand the basic tenets of RQA and will be able to
critically review the literature where RQA is applied
Students will be able to go through all the steps needed for the
analysis, from data preparation, to parameter setting, to presentation
of the results, by using the dedicated packages in the R environment
Assessment criteria
Assessment methods and criteria
Most of the classes will start with a short (3-4 questions) quiz concerning
the material presented in the previous class, and short polls to gauge
students’ confidence and understanding of the current material will be
administered and used to additionally tune the presentation of materials.
These quizzes won’t contribute to the final grade.
Home assignments will contribute to the evaluation and progress made.
There will be 5 home assignments during the course (approximately one
every two classes; 20 points).
Moreover a mid-term and a final exam are envisioned during which
students will solve practical problems using R (mid-term – 30 points) and
performing a full Recurrence Quantification Analysis (final – 50 points).
For these reasons attendance is deemed essential – students are
expected to attend ALL classes, be on time and prepared for discussion.
n general Home assignments will contribute to 20% of the final grade,
Mid-term Exam to 30% and Final Exams for the remaining 50%. Grades
will be assigned according to the following scale:
5 – 90-100% – outstanding performance
4+ – 79-89
4 – 73-78% – good performance
3+ – 67-72
3 – 60-66% – minimum passing performance
2 – 59% or less – performance not suitable for passing
Additional information
Additional information (registration calendar, class conductors, localization and schedules of classes), might be available in the USOSweb system: