Course Overview:
This 5-day course, taught by Dr William Hoppitt, will involve a combination of lectures and practical sessions. Students will learn to build and fit custom models for analysing behavioural data using maximum likelihood techniques in R. This flexible approach allows a researcher to a) use a statistical model that directly represents their hypothesis, in cases where standard models are not appropriate and b) better understand how standard statistical models (e.g. GLMs) are fitted, many of which are fitted by maximum likelihood. Students will learn how to deal with binary, count and continuous data, including time-to-event data which is commonly encountered in behavioural analysis.
After successfully completing this course students should be able to:
- fit a multi-parameter maximum likelihood model in R
- derive likelihood functions for binary, count and continuous data
- deal with time-to-event data
- build custom models to test specific behavioural hypotheses
- conduct hypothesis tests and construct confidence intervals
- use Akaike’s information criterion (AIC) and model averaging
- understand how maximum likelihood relates to Bayesian techniques
Intended Audience
Any researchers (from postgraduate students to senior investigators) interested in analysing behavioural data. Examples will be primarily from non-human animal behaviour studies, but the methods will also be applicable to many researchers studying human behaviour. The course is intended for those wishing to construct custom statistical models and for those wishing to better understand the workings of standard statistical techniques that use maximum likelihood methods (e.g. GLMs).
More information is available from the PS Statistics website.