12.07.2015 Views

Monte Carlo Inference - STAT - EPFL

Monte Carlo Inference - STAT - EPFL

Monte Carlo Inference - STAT - EPFL

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Galaxy dataSpeed10 15 20 25 30 35Normal Q−Q Plot−2 −1 0 1 2Theoretical Quantiles<strong>Monte</strong> <strong>Carlo</strong> <strong>Inference</strong> Spring 2009 – slide 196Mixture density□ Natural model for such data is a p-component mixture densityf(y;θ) =p∑π r f r (y;θ), 0 ≤ π r ≤ 1,r=1p∑π r = 1,where π r is the probability that Y comes from the rth component and f r (y;θ) is its densityconditional on this event.□ Widely used class of models, often with number of components p unknown.□ Aside: such models are non-regular for some likelihood inferences:– non-identifiable under permutation of components;– setting π r = 0 eliminates parameters of f r ;– maximum of likelihood can be +∞, achieved for several θ<strong>Monte</strong> <strong>Carlo</strong> <strong>Inference</strong> Spring 2009 – slide 197r=1197

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!