Modelling egg densities by a zero-inflated Bayesian hierarchical space-time model

• Main Authors: Ingunn Fride Tvete, Lorenzo Ciannelli, Bent Natvig, Geir Storvik
• Other Authors: The Pennsylvania State University CiteSeerX Archives
• Format: Text
• Language: English
• Subjects: Bayesian hierarchical modelling; egg counts; MCMC; space-time modelling; walleye; Gulf of Alaska; Kodiak; Alaska
• Online Access: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.164.6982; http://www.math.uio.no/~geirs/publ/Tvete_etal_v2_2007.pdf

Summary. Data from egg sampling surveys often contain a large proportion of zeros. We examine the time and location of collected walleye pollock egg counts from the western Gulf of Alaska, from Kodiak Island to Unimak Pass, in the years 1978-2000. We take the excessive number of zeros in the data into account by taking a two stage modelling approach, resulting in a zero-inflated hierarchical space-time model. The binary (presence/absence of zeros) process is directly linked to the underlying intensity process. Hence, the latter process influences both the presence of zeros and the amount of eggs in the non-zero observations. We fit our model for each year, and compare the underlying intensities over the years within two regions of the sampled area.