Måns Karlsson: Modelling daily numbers of ringed birds with negative binomial generalized linear models

There is quite an abundance of unanalysed bird population monitoring data in Sweden in general and at Falsterbo Bird Observatory in particular. However, such data is often high variance count data, far from ordinary multiple regression on continuous data. Thus, the number of analyses performed is usually held back by a shortage of people with appropriate skills among researchers and amateurs collecting and owning this data. In this thesis, we will take advantage of the surplus of data and look at the daily number of ringed Eurasian Robins in Falsterbo, trying to predict it through a function of local weather and time. The daily number of Robins is typical bird population monitoring data in the sense that it fits an over dispersed Poisson distribution. In order to account for the higher variance, we will extend the framework of generalized linear models (GLMs) to encompass the negative binomial distribution, which is of great use when dealing with data fitting an over dispersed Poisson distribution.