Hello,
I am exploring options for implementing the model described in Li et al 2012 (
https://doi.org/10.1177/0962280212446326) within INLA. If I understand correctly, the original model was implemented in an MCMC framework to allow for the variable placement of disease incidence in a sampling region using a multinomial distribution, to account for the lack of exact location information.
I was wondering if this step could be similarly achieved by calculating the count per grid cell as described in Li et al (Count/intersections), and then applying the "xpoisson" family instead of the "poisson" to account for the resulting non-integer response? This would similarly spread the potential incidence of disease across the sampling area, but I was unsure if the change in data augmentation/distribution choice would violate assumptions of the LGCP.
I appreciate any guidance, thank you!
Matt