An Extension of the Compound Poisson Information Criterion to at Most Two Change Points
Roach, Jacob S.
MetadataShow full item record
In a data set, there could exist more than one set of parameters. Initially, the observed data could follow a distribution with one set of parameters and at a certain point the parameters of the distribution change. The data points where these changes occur are known as change points. In this thesis we examine the problem of locating change points given observations drawn from randomly generated data as part of a simulation study as well as gene expression data in application. The distribution of interest is a compound Poisson process which is the sum of normal observations. This paper uses an information criterion approach to identify at most two change points. Through simulation and application, this particular problem is examined.