I'm working up to fitting a Tweedie GLMM in R (mgcv). At the moment I'm playing in SPSS to sort out the data and some ideas for random effects. The data are from a survey of coastal dolphins on a sample of sites (Site) around the coast, each measured on a set of transects. The data are the number of groups (NGroups), the group sizes (GSize), and number of individuals (NIndividuals). NIndividuals is the sum of a Poisson number (Ngroups) of Gamma (or Negative Binomial) distributions (Group sizes). Given transect length and the search width, it is possible to calculate the density of sightings for each transect (SightDens). So far I've fitted a Poisson regression to NGroups with a random effect for Site using SPSS Genlinmixed. Now I'd like to see whether there is any spatial autocorrelation among sites. As the sites are strung out around the coast, this can be treated as a purely linear problem based on site order (SiteOrder). I'd like to try to fit a simple AR1 structure to the Site covariance matrix based on SiteOrder. Can anyone help with this?
pls read this paper
https://web.njit.edu/~zhiwei/GAS_101.pdf
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS