GAMM applied on parasite data.

ZUUR A. F.; IENO E. N.; SAVALIEV A. A.; TIMI J. T.; CANTATORE, D. M. P.; HIELBE J. M.

A beginner's guide to generalised additive mixed models with R

Highland Statistics Ltd.

Lugar: Newburgh; Año: 2014; p. 115 - 144

Timi and Poulin (2003) studied the parasite community of the Argentine anchovy (Engraulis anchoita), a small pelagic fish species in the south West Atlantic that plays an important ecological role in the diet of other commercial species such as hake, mackerel and squid. In their publication, Time and Poulin addressed whether parasite communities of anchovies were random collections of species or structured and repetitive sets of species by using nested subset analyses. The relationships between nestedness and community descriptors (i.e. species richness, parasite abundance) and host features (size) were studied by applying nonparametric tests. In this chapter we follow a different approach and we assess the importance of various ecological factors (host body length, sex, latitude and seasonal variation) on different diversity indices. More than 2000 specimens of Engraulis anchoita collected during six cruises were systematically evaluated for parasites and a total of 13 metazoan parasite species were measured. The sampling program consisted of midwater trawl nets and covered four different populations: the north autumn Bonaerense and three spring populations, northBonaerense, south Bonaerense and Patagonian. Various diversity indices are described in Magurran (2004), and common used indices are total abundance, species richness, Shannon Wiener, absence/presence, Simpson index, Berger-Parker index, among many others. In a paper or report one would normally use only one of these indices, but in this chapter we use total abundance, species richnessand absence/presence. The reason for this is that, as we will see later, thetotal abundance requires a negative binomial generalized additive mixedeffects model (GAMM), species richness requires a generalized Poisson GAMM to deal with under-dispersion, and the absence/presence data requires a binomial GAMM. That means that in one chapter we can explain 3 different types of GAMM using a relatively simple-tounderstand data set.