Event Date
This is a ‘LIVE COURSE’ – the instructor will be delivering lectures and coaching attendees
through the accompanying computer practicals via video link, so a good internet connection is
essential.
TIME ZONE – Western European Time – however all sessions will be recorded and made available allowing attendees from different time zones to follow. However, all sessions will be recorded and made available, allowing attendees from different time zones to follow asynchronously.
Please email oliverhooker@prstatistics.com for full details or to discuss how we can accommodate you).
Hidden Markov models (HMMs) are flexible statistical models for time series observations driven by underlying states. Over the last decade, HMMs have become increasingly popular within the ecological community as they allow to uncover behavioural state dynamics from noisy sensor data. For example, a typical HMM-based analysis of say GPS locations or acceleration measurements could involve the investigation of internal (e.g. sex, size, age) and external (e.g. temperature, habitat) drivers of behavioural state occupancy.
This workshop will introduce the HMM framework, comprising a mix of theoretical lectures and hands-on practical components using R. In the theoretical sessions, the following topics will be covered:
motivation & overview
basic HMM formulation
fitting an HMM to data
model selection & model checking
state decoding
incorporating covariates, seasonality and random effects
other model extensions
These techniques will be illustrated primarily using movement and acceleration data, but are applicable also to other ecological time series data (e.g. capture-recapture). In the practical sessions, we will focus on HMM analyses using the R packages moveHMM and momentuHMM, but will also showcase the use of hmmTMB. Basic knowledge of the free software R is helpful, but not required.
A basic understanding of statistics and probability calculus, as it would be taught in any introductory statistics class, is required. By the end of the course, participants will have a good understanding of what HMMs are and what they can be used for. Participants will also be prepared to tailor a suitable HMM to their data and to implement the corresponding analysis in R.
Academics and post-graduate students interested in adding HMMs to their methodological toolbox for analysing ecological data.
Delivered remotely
Availability – 25 places
Duration – 3.5 days
Contact hours – Approx. 24 hours
ECT’s – Equal to 2 ECT’s
Language – English
We will spend the mornings to learn about HMM methodology, but will make these lectures as
interactive as possible, with several R code snippets to try out and lots of time for questions. In the
afternoons, we will implement example HMM analyses in R. Some example data sets for the practical sessions will be provided by the instructors, but participants are welcome to bring their own data. A Slack channel will be open to discuss any issues for which we may not have enough time in the sessions themselves.
Basic understanding of statistics and probability calculus (e.g. probability distributions, density functions, conditional probability).
Basic familiarity with R is sufficient. In fact, participants will be able to follow most of the workshop
without prior knowledge of R.
Participants must use a computer with a good internet connection, a working recent version or R (and ideally also RStudio), and recent versions of some R packages whose installation instructions will be sent a few days before the course. A working webcam is desirable for enhanced interactivity during the live sessions. Some computation power is required for modelling large datasets, although the provided example data (and suggested subsets of participants’ data) can run on an ordinary laptop.
Cancellations are accepted up to 28 days before the course start date subject to a 25% cancellation fee. Cancellations later than this may be considered, contact oliverhooker@prstatistics.com. Failure to attend will result in the full cost of the course being charged. In the unfortunate event that a course is cancelled due to unforeseen circumstances a full refund of the course fees will be credited.
If you are unsure about course suitability, please get in touch by email to find out more oliverhooker@prstatistics.com
Day 1 – Classes from 08:30 – 16:00
Three 1-hour theory sessions:
motivation & overview
preliminaries: probability calculus & Markov chains
the basic HMM formulation
A 1-hour practical session:
simulating data from an HMM
Day 2 – Classes from 08:30 – 16:00
Three 1-hour theory sessions:
fitting an HMM to real data, part I
fitting an HMM to real data, part II
fitting HMMs to movement and acceleration data
A 2-hour practical session:
fitting an HMM to real data
Day 3 – Classes from 08:30 – 16:00
Three 1-hour theory sessions:
model selection & model checking
state decoding
covariates
A 2-hour practical session:
complete HMM workflow
Day 4 – Classes from 08:30 – 11:00
Three 1-hour theory sessions:
overview of extensions, part I
overview of extensions, part II
time discuss participants’ own data/questions
Roland is a professor of statistics and data analysis at Bielefeld University in Germany, where he is heavily involved in the teaching of introductory statistics courses as well as advanced statistical methods. His research focuses on statistical method development for state-switching time series models, in particular hidden Markov models, as well as their applications primarily in ecology, sports and economics. Within statistical ecology, he has published extensively on the modelling of animal movement and general behaviour, but also on capture-recapture and distance sampling.
Google Scholar
Homepage