Chapter 11 Exercises
Mixture models for eye tracking
Solutions
In the monkey eye tracking example 11.6.1:
a) Rewrite the BUGS code for the mixture model using the Bernoulli distribution dbern() instead of dcat(), and check the results are the same.
b) Extend the mixture model so that the variances of each mixture component are different.
c) (advanced) By rewriting using an explicit likelihood, it should be possible to compare the DIC of this model with the common-variance model. However it is difficult to find a parameterisation and prior structure for which the sampling model converges, as illustrated in the solutions. If you find a solution, please get in touch with the authors!
a)
click arrow for dbern() implementation
b)
click arrow for different-variances model
c)
click for different-variances model using an explicit likelihood (doesn't work!)