Inference in a Partially Observed Percolation Process

• In this licentiate thesis, inference in a partially oberved percolation process living on a graph, is considered. Each edge of the graph is declared open with probability $\theta$ and closed with probability $1-\theta$ independently of the states of all other edges. The inference problem is to draw inference about $\theta$ based on the information on whether or not particular pairs of vertices are connected by open paths. Consistency results under certain conditions on the graph are given for both a Bayesian and a frequentist approach to the inference problem. Moreover, a simulation study is presented which in addition to illustrating the consistency results, also indicates that the consistency results might hold for percolation processes on more general graphs.

Subject headings

NATURVETENSKAP  -- Matematik -- Sannolikhetsteori och statistik (hsv//swe)

NATURAL SCIENCES  -- Mathematics -- Probability Theory and Statistics (hsv//eng)

Keyword

Percolation

Bayesian inference

frequentistic inference

consistency

Markov chain Monte Carlo

Monte Carlo Expectation Maximization

Markov chain Monte Carlo

Publication and Content Type

vet (subject category)

lic (subject category)