skip to content

Inference with approximate likelihoods

Presented by: 
Helen Ogden
Monday 3rd July 2017 - 15:30 to 16:15
INI Seminar Room 1
In cases where it is infeasible to compute the likelihood exactly, an alternative is to find some numerical approximation to the likelihood, then to use this approximate likelihood in place of the exact likelihood to do inference about the model parameters. This is a fairly commonly used approach, and I will give several examples of approximate likelihoods which have been used in this way. But is this a valid approach to inference? I will give conditions under which inference with an approximate likelihood shares some of the same asymptotic properties as inference with the exact likelihood, and describe the implications in some examples. I will finish with some ideas about how to construct scalable likelihood approximations which give statistically valid inference.
The video for this talk should appear here if JavaScript is enabled.
If it doesn't, something may have gone wrong with our embedded player.
We'll get it fixed as soon as possible.
University of Cambridge Research Councils UK
    Clay Mathematics Institute London Mathematical Society NM Rothschild and Sons